jettask 0.2.19__py3-none-any.whl → 0.2.20__py3-none-any.whl

jettask/core/app.py

@@ -1,28 +1,39 @@
 import os
-import sys
+
 import time
 from datetime import datetime
 from ..utils.serializer import dumps, loads, dumps_str, loads_str
 import signal
-import socket
+
 import asyncio
 import logging
 import contextlib
 import importlib
-import multiprocessing
+import time 
+
 from typing import List
-from collections import defaultdict, deque
+
 
 import redis
 from redis import asyncio as aioredis
-from watchdog.observers import Observer
 
+# 导入TaskMessage
+from .message import TaskMessage
 from .task import Task
-from .event_pool import EventPool
-from ..executors import AsyncioExecutor, MultiAsyncioExecutor
-from ..monitoring import FileChangeHandler
+from .enums import TaskStatus
+from jettask.messaging.event_pool import EventPool
+from ..executor.orchestrator import ProcessOrchestrator
 from ..utils import gen_task_name
 from ..exceptions import TaskTimeoutError, TaskExecutionError, TaskNotFoundError
+# 导入统一的数据库连接管理
+from ..utils.db_connector import get_sync_redis_client, get_async_redis_client
+# 导入Lua脚本
+from ..config.lua_scripts import (
+    LUA_SCRIPT_DELAYED_TASKS,
+    LUA_SCRIPT_NORMAL_TASKS,
+    LUA_SCRIPT_SEND_DELAYED_TASKS
+)
+import uvloop
 
 logger = logging.getLogger('app')
 
@@ -33,251 +44,59 @@ logging.basicConfig(
     datefmt="%Y-%m-%d %H:%M:%S",
 )
 
-# 尝试导入性能优化库
-try:
-    import uvloop
-    UVLOOP_AVAILABLE = True
-    # 自动启用uvloop
-    asyncio.set_event_loop_policy(uvloop.EventLoopPolicy())
-    logger.debug("Using uvloop for better performance")
-except ImportError:
-    UVLOOP_AVAILABLE = False
-
-_on_app_finalizers = set()
-
-# 全局连接池复用
-_redis_pools = {}
-_async_redis_pools = {}
-# 专门用于二进制数据的连接池（用于Stream操作）
-_binary_redis_pools = {}
-_async_binary_redis_pools = {}
-
-def get_redis_pool(redis_url: str, max_connections: int = 200):
-    """获取或创建Redis连接池"""
-    if redis_url not in _redis_pools:
-        # 构建socket keepalive选项，仅在Linux上使用
-        socket_keepalive_options = {}
-        if hasattr(socket, 'TCP_KEEPIDLE'):
-            socket_keepalive_options[socket.TCP_KEEPIDLE] = 1
-        if hasattr(socket, 'TCP_KEEPINTVL'):
-            socket_keepalive_options[socket.TCP_KEEPINTVL] = 3
-        if hasattr(socket, 'TCP_KEEPCNT'):
-            socket_keepalive_options[socket.TCP_KEEPCNT] = 5
-        
-        _redis_pools[redis_url] = redis.ConnectionPool.from_url(
-            redis_url, 
-            decode_responses=True,
-            max_connections=max_connections,
-            retry_on_timeout=True,
-            retry_on_error=[ConnectionError, TimeoutError],
-            socket_keepalive=True,
-            socket_keepalive_options=socket_keepalive_options if socket_keepalive_options else None,
-            health_check_interval=30,
-            # 优化超时配置以处理高负载
-            socket_connect_timeout=10,  # 增加连接超时时间
-            socket_timeout=15,          # 增加读取超时时间，避免频繁超时
-        )
-    return _redis_pools[redis_url]
-
-def get_async_redis_pool(redis_url: str, max_connections: int = 200):
-    """获取或创建异步Redis连接池"""
-    if redis_url not in _async_redis_pools:
-        # 构建socket keepalive选项，仅在Linux上使用
-        socket_keepalive_options = {}
-        if hasattr(socket, 'TCP_KEEPIDLE'):
-            socket_keepalive_options[socket.TCP_KEEPIDLE] = 1
-        if hasattr(socket, 'TCP_KEEPINTVL'):
-            socket_keepalive_options[socket.TCP_KEEPINTVL] = 3
-        if hasattr(socket, 'TCP_KEEPCNT'):
-            socket_keepalive_options[socket.TCP_KEEPCNT] = 5
-        
-        _async_redis_pools[redis_url] = aioredis.ConnectionPool.from_url(
-            redis_url,
-            decode_responses=True,
-            max_connections=max_connections,
-            retry_on_timeout=True,
-            retry_on_error=[ConnectionError, TimeoutError],
-            socket_keepalive=True,
-            socket_keepalive_options=socket_keepalive_options if socket_keepalive_options else None,
-            health_check_interval=30,
-            # 优化超时配置以处理高负载
-            socket_connect_timeout=10,  # 增加连接超时时间
-            socket_timeout=15,          # 增加读取超时时间，避免频繁超时
-        )
-    return _async_redis_pools[redis_url]
-
-def get_binary_redis_pool(redis_url: str, max_connections: int = 200):
-    """获取或创建用于二进制数据的Redis连接池（Stream操作需要）"""
-    if redis_url not in _binary_redis_pools:
-        # 构建socket keepalive选项，仅在Linux上使用
-        socket_keepalive_options = {}
-        if hasattr(socket, 'TCP_KEEPIDLE'):
-            socket_keepalive_options[socket.TCP_KEEPIDLE] = 1
-        if hasattr(socket, 'TCP_KEEPINTVL'):
-            socket_keepalive_options[socket.TCP_KEEPINTVL] = 3
-        if hasattr(socket, 'TCP_KEEPCNT'):
-            socket_keepalive_options[socket.TCP_KEEPCNT] = 5
-        
-        _binary_redis_pools[redis_url] = redis.ConnectionPool.from_url(
-            redis_url, 
-            decode_responses=False,  # 不解码，因为Stream数据是msgpack二进制
-            max_connections=max_connections,
-            retry_on_timeout=True,
-            retry_on_error=[ConnectionError, TimeoutError],
-            socket_keepalive=True,
-            socket_keepalive_options=socket_keepalive_options if socket_keepalive_options else None,
-            health_check_interval=30,
-            # 优化超时配置以处理高负载
-            socket_connect_timeout=10,  # 增加连接超时时间
-            socket_timeout=15,          # 增加读取超时时间，避免频繁超时
-        )
-    return _binary_redis_pools[redis_url]
-
-def get_async_binary_redis_pool(redis_url: str, max_connections: int = 200):
-    """获取或创建用于二进制数据的异步Redis连接池（Stream操作需要）"""
-    if redis_url not in _async_binary_redis_pools:
-        # 构建socket keepalive选项，仅在Linux上使用
-        socket_keepalive_options = {}
-        if hasattr(socket, 'TCP_KEEPIDLE'):
-            socket_keepalive_options[socket.TCP_KEEPIDLE] = 1
-        if hasattr(socket, 'TCP_KEEPINTVL'):
-            socket_keepalive_options[socket.TCP_KEEPINTVL] = 3
-        if hasattr(socket, 'TCP_KEEPCNT'):
-            socket_keepalive_options[socket.TCP_KEEPCNT] = 5
-        
-        _async_binary_redis_pools[redis_url] = aioredis.ConnectionPool.from_url(
-            redis_url,
-            decode_responses=False,  # 不解码，因为Stream数据是msgpack二进制
-            max_connections=max_connections,
-            retry_on_timeout=True,
-            retry_on_error=[ConnectionError, TimeoutError],
-            socket_keepalive=True,
-            socket_keepalive_options=socket_keepalive_options if socket_keepalive_options else None,
-            health_check_interval=30,
-            # 优化超时配置以处理高负载
-            socket_connect_timeout=10,  # 增加连接超时时间
-            socket_timeout=15,          # 增加读取超时时间，避免频繁超时
-        )
-    return _async_binary_redis_pools[redis_url]
-
-
-def connect_on_app_finalize(callback):
-    """Connect callback to be called when any app is finalized."""
-    _on_app_finalizers.add(callback)
-    return callback
+UVLOOP_AVAILABLE = True
+# 自动启用uvloop
+asyncio.set_event_loop_policy(uvloop.EventLoopPolicy())
+logger.debug("Using uvloop for better performance")
 
 
 class Jettask(object):
-    # Lua脚本定义为类常量，避免重复定义
-    _LUA_SCRIPT_DELAYED_TASKS = """
-    local prefix = ARGV[1]
-    local current_time = tonumber(ARGV[2])
-    local results = {}
-    
-    -- 从ARGV[3]开始，每5个参数为一组任务信息
-    -- [stream_key, stream_data, execute_at, delay_seconds, queue]
-    for i = 3, #ARGV, 5 do
-        local stream_key = ARGV[i]
-        local stream_data = ARGV[i+1]
-        local execute_at = tonumber(ARGV[i+2])
-        local delay_seconds = tonumber(ARGV[i+3])
-        local queue = ARGV[i+4]
-        
-        -- 使用Hash存储所有队列的offset
-        local offsets_hash = prefix .. ':QUEUE_OFFSETS'
-        -- 使用HINCRBY原子递增offset
-        local offset = redis.call('HINCRBY', offsets_hash, queue, 1)
-        
-        -- 1. 添加消息到Stream（包含offset字段）
-        local stream_id = redis.call('XADD', stream_key, '*', 
-            'data', stream_data,
-            'offset', offset)
-        
-        -- 2. 添加到延迟队列ZSET
-        local delayed_queue_key = prefix .. ':DELAYED_QUEUE:' .. queue
-        redis.call('ZADD', delayed_queue_key, execute_at, stream_id)
-        
-        -- 3. 设置任务状态Hash（只存储status，其他信息从Stream获取）
-        local task_key = prefix .. ':TASK:' .. stream_id
-        redis.call('HSET', task_key, 'status', 'delayed')
-        
-        -- 4. 设置过期时间
-        local expire_seconds = math.max(1, math.floor(delay_seconds + 3600))
-        redis.call('EXPIRE', task_key, expire_seconds)
-        
-        -- 保存stream_id到结果
-        table.insert(results, stream_id)
-    end
-    
-    return results
-    """
-    
-    _LUA_SCRIPT_NORMAL_TASKS = """
-    local prefix = ARGV[1]
-    local current_time = ARGV[2]
-    local results = {}
-    
-    -- 从ARGV[3]开始，每2个参数为一组任务信息
-    -- [stream_key, stream_data]
-    for i = 3, #ARGV, 2 do
-        local stream_key = ARGV[i]
-        local stream_data = ARGV[i+1]
-        
-        -- 从stream_key中提取队列名（格式: prefix:STREAM:queue_name）
-        local queue_name = string.match(stream_key, prefix .. ':STREAM:(.*)')
-        
-        -- 获取并递增offset
-        local offset_key = prefix .. ':STREAM:' .. queue_name .. ':next_offset'
-        local offset = redis.call('INCR', offset_key)
-        
-        -- 1. 添加消息到Stream（包含offset字段）
-        local stream_id = redis.call('XADD', stream_key, '*', 
-            'data', stream_data,
-            'offset', offset)
-        
-        -- 2. 设置任务状态Hash（只存储status）
-        local task_key = prefix .. ':TASK:' .. stream_id
-        redis.call('HSET', task_key, 'status', 'pending')
-        
-        -- 3. 设置过期时间（1小时）
-        redis.call('EXPIRE', task_key, 3600)
-        
-        -- 保存stream_id到结果
-        table.insert(results, stream_id)
-    end
-    
-    return results
-    """
+    # Lua脚本从config模块导入，统一管理
+    _LUA_SCRIPT_DELAYED_TASKS = LUA_SCRIPT_DELAYED_TASKS
+    _LUA_SCRIPT_NORMAL_TASKS = LUA_SCRIPT_NORMAL_TASKS
 
-    def __init__(self, redis_url: str = None, include: list = None, max_connections: int = 200, 
+    def __init__(self, redis_url: str = None, include: list = None, max_connections: int = None,
                  consumer_strategy: str = None, consumer_config: dict = None, tasks=None,
                  redis_prefix: str = None, scheduler_config: dict = None, pg_url: str = None,
-                 task_center=None) -> None:
+                 task_center=None, worker_id: str = None, worker_key: str = None) -> None:
         self._tasks = tasks or {}
         self._queue_tasks = {}  # 记录每个队列对应的任务列表
         self.asyncio = False
         self.include = include or []
-        
+
         # 任务中心相关属性
         self.task_center = None  # 将通过mount_task_center方法挂载或初始化时指定
         self._task_center_config = None
         self._original_redis_url = redis_url
         self._original_pg_url = pg_url
-        
-        self.redis_url = redis_url
-        self.pg_url = pg_url  # 存储PostgreSQL URL
-        self.max_connections = max_connections
+
+        # 优先使用传入参数，其次使用环境变量
+        self.redis_url = redis_url or os.environ.get('JETTASK_REDIS_URL')
+        self.pg_url = pg_url or os.environ.get('JETTASK_PG_URL')
+        self.max_connections = max_connections if max_connections is not None else int(os.environ.get('JETTASK_MAX_CONNECTIONS', '500'))
+        self.redis_prefix = redis_prefix or os.environ.get('JETTASK_REDIS_PREFIX', 'jettask')
+
+        # 检查必需参数：redis_url
+        if not self.redis_url:
+            raise ValueError(
+                "必须提供 redis_url 参数！\n\n"
+                "请通过以下任一方式配置:\n"
+                "  1. 初始化时传参:\n"
+                "     app = Jettask(redis_url='redis://localhost:6379/0')\n\n"
+                "  2. 设置环境变量:\n"
+                "     export JETTASK_REDIS_URL='redis://localhost:6379/0'\n\n"
+                "  3. 在 .env 文件中配置:\n"
+                "     JETTASK_REDIS_URL=redis://localhost:6379/0\n"
+            )
+
         self.consumer_strategy = consumer_strategy
         self.consumer_config = consumer_config or {}
         self.scheduler_config = scheduler_config or {}
-        
-        # Redis prefix configuration
-        self.redis_prefix = redis_prefix or "jettask"
-        
+
         # 如果初始化时提供了task_center，直接挂载
         if task_center:
             self.mount_task_center(task_center)
-        
+
         # Update prefixes with the configured prefix using colon namespace
         self.STATUS_PREFIX = f"{self.redis_prefix}:STATUS:"
         self.RESULT_PREFIX = f"{self.redis_prefix}:RESULT:"
@@ -289,15 +108,33 @@ class Jettask(object):
         # 调度器相关
         self.scheduler = None
         self.scheduler_manager = None
-        
+
         self._status_prefix = self.STATUS_PREFIX
         self._result_prefix = self.RESULT_PREFIX
-        
+
+        # Worker 状态管理器（延迟初始化）
+        self.worker_state_manager = None
+
+        # Worker 状态查询器（延迟初始化 - 需要 Redis 客户端）
+        self._worker_state = None
+
+        # Worker ID（可选，用于子进程复用主进程的 ID）
+        self.worker_id = worker_id
+        self.worker_key = worker_key
+
         # 初始化清理状态，但不注册处理器
         self._cleanup_done = False
         self._should_exit = False
         self._worker_started = False
         self._handlers_registered = False
+
+        # 初始化队列注册表（用于获取任务名称等操作）
+        from ..messaging.registry import QueueRegistry
+        self.registry = QueueRegistry(
+            redis_client=None,  # 延迟初始化，第一次使用时通过 self.redis 获取
+            async_redis_client=None,
+            redis_prefix=self.redis_prefix
+        )
    
     
     def _load_config_from_task_center(self):
@@ -343,7 +180,7 @@ class Jettask(object):
                     else:
                         self.redis_url = f"redis://{redis_host}:{redis_port}/{redis_db}"
                     
-                    logger.info(f"从任务中心加载Redis配置: {redis_host}:{redis_port}/{redis_db}")
+                    logger.debug(f"从任务中心加载Redis配置: {redis_host}:{redis_port}/{redis_db}")
                 
                 # 构建PostgreSQL URL
                 if pg_config:
@@ -354,7 +191,7 @@ class Jettask(object):
                     pg_database = pg_config.get('database', 'jettask')
                     
                     self.pg_url = f"postgresql://{pg_user}:{pg_password}@{pg_host}:{pg_port}/{pg_database}"
-                    logger.info(f"从任务中心加载PostgreSQL配置: {pg_host}:{pg_port}/{pg_database}")
+                    logger.debug(f"从任务中心加载PostgreSQL配置: {pg_host}:{pg_port}/{pg_database}")
                 
                 # 保存配置供后续使用
                 self._task_center_config = config
@@ -392,7 +229,7 @@ class Jettask(object):
             task_center: TaskCenter实例
             
         使用示例：
-            from jettask.task_center import TaskCenter
+            from jettask.task.task_center.client import TaskCenter
             
             # 创建任务中心客户端（可复用）
             task_center = TaskCenter("http://localhost:8001/api/namespaces/demo")
@@ -456,7 +293,7 @@ class Jettask(object):
                 return
             # 只有启动过worker才需要打印清理信息
             if self._worker_started:
-                logger.info("Received shutdown signal, cleaning up...")
+                logger.debug("Received shutdown signal, cleaning up...")
             self.cleanup()
             if signum:
                 # 设置标记表示需要退出
@@ -485,23 +322,22 @@ class Jettask(object):
         """清理应用资源"""
         if self._cleanup_done:
             return
-        
         self._cleanup_done = True
         
         # 只有真正启动过worker才打印日志
         if self._worker_started:
-            logger.info("Cleaning up Jettask resources...")
+   
             
             # 清理EventPool
             if hasattr(self, 'ep') and self.ep:
                 self.ep.cleanup()
             
-            logger.info("Jettask cleanup completed")
+
         else:
             # 如果只是实例化但没有启动，静默清理
             if hasattr(self, 'ep') and self.ep:
                 self.ep.cleanup()
-            logger.debug("Jettask instance cleanup (no worker started)")
+
     
     @property
     def consumer_manager(self):
@@ -510,35 +346,67 @@ class Jettask(object):
 
     @property
     def async_redis(self):
-        """优化：复用连接池"""
-        name = "_async_redis"
-        if hasattr(self, name):
-            return getattr(self, name)
-        
+        """获取异步Redis客户端（全局单例）"""
         # 如果配置了任务中心且还未加载配置，先加载配置
         if self.task_center and self.task_center.is_enabled and not self._task_center_config:
             self._load_config_from_task_center()
-        
-        pool = get_async_redis_pool(self.redis_url, self.max_connections)
-        async_redis = aioredis.StrictRedis(connection_pool=pool)
-        setattr(self, name, async_redis)
-        return async_redis
+
+        # 使用超长超时时间，支持 Pub/Sub 长连接（可能几天没有消息）
+        return get_async_redis_client(
+            redis_url=self.redis_url,
+            decode_responses=True,
+            max_connections=self.max_connections,
+            socket_timeout=99999999999  # 超长超时，几乎永不超时
+        )
 
     @property
     def redis(self):
-        """优化：复用连接池"""
-        name = "_redis"
-        if hasattr(self, name):
-            return getattr(self, name)
-        
+        """获取同步Redis客户端（全局单例）"""
         # 如果配置了任务中心且还未加载配置，先加载配置
         # if self.task_center and self.task_center.is_enabled and not self._task_center_config:
         #     self._load_config_from_task_center()
-        print(f'{self.redis_url=}')
-        pool = get_redis_pool(self.redis_url, self.max_connections)
-        redis_cli = redis.StrictRedis(connection_pool=pool)
-        setattr(self, name, redis_cli)
-        return redis_cli
+
+        return get_sync_redis_client(
+            redis_url=self.redis_url,
+            decode_responses=True,
+            max_connections=self.max_connections
+        )
+
+    @property
+    def binary_redis(self):
+        """获取同步二进制Redis客户端（不自动解码，用于获取msgpack数据）"""
+        return get_sync_redis_client(
+            redis_url=self.redis_url,
+            decode_responses=False,
+            max_connections=self.max_connections
+        )
+
+    @property
+    def async_binary_redis(self):
+        """获取异步二进制Redis客户端（不自动解码，用于获取msgpack数据）"""
+        return get_async_redis_client(
+            redis_url=self.redis_url,
+            decode_responses=False,
+            max_connections=self.max_connections,
+            socket_timeout=99999999999
+        )
+
+    @property
+    def worker_state(self):
+        """
+        获取 WorkerState 实例（单例，延迟初始化）
+
+        WorkerState 负责 Worker 状态的查询和管理
+        """
+        if self._worker_state is None:
+            from jettask.worker.manager import WorkerState
+            self._worker_state = WorkerState(
+                redis_client=self.redis,
+                async_redis_client=self.async_redis,
+                redis_prefix=self.redis_prefix
+            )
+            logger.debug("Initialized WorkerState for app")
+        return self._worker_state
 
     @property
     def ep(self):
@@ -549,10 +417,10 @@ class Jettask(object):
             # 传递redis_prefix到consumer_config
             consumer_config = self.consumer_config.copy() if self.consumer_config else {}
             consumer_config['redis_prefix'] = self.redis_prefix
-            
+
             ep = EventPool(
-                self.redis, 
-                self.async_redis, 
+                self.redis,
+                self.async_redis,
                 redis_url=self.redis_url,
                 consumer_strategy=self.consumer_strategy,
                 consumer_config=consumer_config,
@@ -592,13 +460,13 @@ class Jettask(object):
         self.include += modules
 
     def _task_from_fun(
-        self, fun, name=None, base=None, queue=None, bind=False, retry_config=None, **options
+        self, fun, name=None, base=None, queue=None, bind=False, retry_config=None, rate_limit=None, **options
     ) -> Task:
         name = name or gen_task_name(fun.__name__, fun.__module__)
         base = base or Task
-        
+
         # 不再限制队列模式，因为每个task都有独立的consumer group
-        
+
         if name not in self._tasks:
             run = staticmethod(fun)
             task: Task = type(
@@ -611,6 +479,7 @@ class Jettask(object):
                         "run": run,
                         "queue": queue,
                         "retry_config": retry_config,  # 存储重试配置
+                        "rate_limit": rate_limit,  # 存储限流配置
                         "_decorated": True,
                         "__doc__": fun.__doc__,
                         "__module__": fun.__module__,
@@ -624,16 +493,67 @@ class Jettask(object):
             with contextlib.suppress(AttributeError):
                 task.__qualname__ = fun.__qualname__
             self._tasks[task.name] = task
-            
+
             # 记录队列和任务的映射（用于查找）
             if queue:
                 if queue not in self._queue_tasks:
                     self._queue_tasks[queue] = []
                 self._queue_tasks[queue].append(name)
+
+            # 如果任务配置了限流，注册到Redis；否则删除旧配置
+            if rate_limit:
+                # 支持 int (QPS) 和 ConcurrencyLimit/QPSLimit 对象
+                if isinstance(rate_limit, int) and rate_limit > 0:
+                    # 简单的 int 值作为 QPS 限制
+                    self._register_rate_limit(name, rate_limit)
+                elif hasattr(rate_limit, 'to_dict'):
+                    # RateLimitConfig 对象（ConcurrencyLimit 或 QPSLimit）
+                    self._register_rate_limit_config(name, rate_limit)
+            else:
+                # 没有限流配置，删除 Redis 中的旧配置（如果存在）
+                from jettask.utils.rate_limit.limiter import RateLimiterManager
+
+                RateLimiterManager.unregister_rate_limit_config(
+                    redis_client=self.redis,
+                    task_name=name,
+                    redis_prefix=self.redis_prefix
+                )
         else:
             task = self._tasks[name]
         return task
-    
+
+    def _register_rate_limit(self, task_name: str, qps_limit: int):
+        """注册任务的 QPS 限流规则到 Redis"""
+        from jettask.utils.rate_limit.config import QPSLimit
+        from jettask.utils.rate_limit.limiter import RateLimiterManager
+
+        # 转换为 QPSLimit 配置对象
+        config = QPSLimit(qps=qps_limit)
+        # 调用 limiter.py 中的静态方法
+        RateLimiterManager.register_rate_limit_config(
+            redis_client=self.redis,
+            task_name=task_name,
+            config=config,
+            redis_prefix=self.redis_prefix
+        )
+
+    def _register_rate_limit_config(self, task_name: str, config):
+        """注册任务的限流配置对象到 Redis
+
+        Args:
+            task_name: 任务名称
+            config: RateLimitConfig 对象（ConcurrencyLimit 或 QPSLimit）
+        """
+        from jettask.utils.rate_limit.limiter import RateLimiterManager
+
+        # 调用 limiter.py 中的静态方法
+        RateLimiterManager.register_rate_limit_config(
+            redis_client=self.redis,
+            task_name=task_name,
+            config=config,
+            redis_prefix=self.redis_prefix
+        )
+
     def task(
         self,
         name: str = None,
@@ -644,6 +564,8 @@ class Jettask(object):
         retry_backoff: bool = True,  # 是否使用指数退避
         retry_backoff_max: float = 60,  # 最大退避时间（秒）
         retry_on_exceptions: tuple = None,  # 可重试的异常类型
+        # 限流相关参数
+        rate_limit: int = None,  # QPS 限制（每秒允许执行的任务数）
         *args,
         **kwargs,
     ):
@@ -659,56 +581,113 @@ class Jettask(object):
                 # 将异常类转换为类名字符串，以便序列化
                 if retry_on_exceptions:
                     retry_config['retry_on_exceptions'] = [
-                        exc if isinstance(exc, str) else exc.__name__ 
+                        exc if isinstance(exc, str) else exc.__name__
                         for exc in retry_on_exceptions
                     ]
-            return self._task_from_fun(fun, name, base, queue, retry_config=retry_config, *args, **kwargs)
+            return self._task_from_fun(fun, name, base, queue, retry_config=retry_config, rate_limit=rate_limit, *args, **kwargs)
 
         return _create_task_cls
     
-    async def send_tasks(self, messages: list):
+    def include_router(self, router, prefix: str = None):
         """
-        统一的任务发送接口 - 只有这一个发送方法
+        包含一个TaskRouter，将其所有任务注册到app中
         
         Args:
-            messages: TaskMessage对象列表（或字典列表）
+            router: TaskRouter实例
+            prefix: 额外的前缀（可选）
+        """
+        from ..task.router import TaskRouter
+        
+        if not isinstance(router, TaskRouter):
+            raise TypeError(f"Expected TaskRouter, got {type(router)}")
+        
+        # 获取router中的所有任务
+        tasks = router.get_tasks()
+        
+        for task_name, task_config in tasks.items():
+            # 复制配置，避免修改原始数据
+            config = task_config.copy()
             
+            # 如果指定了额外前缀，添加到任务名前面
+            if prefix:
+                if config.get('name'):
+                    config['name'] = f"{prefix}.{config['name']}"
+                else:
+                    config['name'] = f"{prefix}.{task_name}"
+            
+            # 获取任务函数和配置
+            func = config.pop('func')
+            name = config.pop('name', task_name)
+            queue = config.pop('queue', None)
+            
+            # 提取重试相关参数
+            retry_config = {}
+            if 'max_retries' in config:
+                retry_config['max_retries'] = config.pop('max_retries', 0)
+            if 'retry_delay' in config:
+                retry_config['retry_backoff_max'] = config.pop('retry_delay', 60)
+            
+            # 注册任务到app
+            self._task_from_fun(
+                func,
+                name=name,
+                queue=queue,
+                retry_config=retry_config if retry_config else None,
+                **config
+            )
+    
+    def send_tasks(self, messages: list, asyncio: bool = False):
+        """
+        统一的任务发送接口 - 支持同步和异步
+
+        Args:
+            messages: TaskMessage对象列表（或字典列表）
+            asyncio: 是否使用异步模式（默认False）
+
         Returns:
-            List[str]: 任务ID列表
-            
+            同步模式: List[str] - 任务ID列表
+            异步模式: 返回协程，需要使用 await
+
         使用示例:
             from jettask.core.message import TaskMessage
-            
-            # 发送单个任务（也是用列表）
+
+            # 同步发送
             msg = TaskMessage(
                 queue="order_processing",
                 args=(12345,),
                 kwargs={"customer_id": "C001", "amount": 99.99}
             )
-            task_ids = await app.send_tasks([msg])
-            
+            task_ids = app.send_tasks([msg])
+
+            # 异步发送
+            task_ids = await app.send_tasks([msg], asyncio=True)
+
             # 批量发送
             messages = [
                 TaskMessage(queue="email", kwargs={"to": "user1@example.com"}),
                 TaskMessage(queue="email", kwargs={"to": "user2@example.com"}),
                 TaskMessage(queue="sms", kwargs={"phone": "123456789"}),
             ]
-            task_ids = await app.send_tasks(messages)
-            
+            task_ids = app.send_tasks(messages)
+
             # 跨项目发送（不需要task定义）
             messages = [
                 TaskMessage(queue="remote_queue", kwargs={"data": "value"})
             ]
-            task_ids = await app.send_tasks(messages)
+            task_ids = await app.send_tasks(messages, asyncio=True)
         """
+        if asyncio:
+            return self._send_tasks_async(messages)
+        else:
+            return self._send_tasks_sync(messages)
+
+    def _send_tasks_sync(self, messages: list):
+        """同步发送任务"""
         if not messages:
             return []
-        
-        # 导入TaskMessage
-        from .message import TaskMessage
-        
+
         results = []
-        
+
         # 按队列分组消息，以便批量处理
         queue_messages = {}
         for msg in messages:
@@ -717,10 +696,10 @@ class Jettask(object):
                 msg = TaskMessage.from_dict(msg)
             elif not isinstance(msg, TaskMessage):
                 raise ValueError(f"Invalid message type: {type(msg)}. Expected TaskMessage or dict")
-            
+
             # 验证消息
             msg.validate()
-            
+
             # 确定实际的队列名（考虑优先级）
             actual_queue = msg.queue
             if msg.priority is not None:
@@ -728,32 +707,69 @@ class Jettask(object):
                 actual_queue = f"{msg.queue}:{msg.priority}"
                 # 更新消息体中的queue字段，确保与实际发送的stream key一致
                 msg.queue = actual_queue
-            
+
             # 按队列分组
             if actual_queue not in queue_messages:
                 queue_messages[actual_queue] = []
             queue_messages[actual_queue].append(msg)
-        
+
         # 处理每个队列的消息
         for queue, queue_msgs in queue_messages.items():
-            # 统一使用批量发送，无论是否广播模式
-            # 广播/单播由消费端的consumer group name决定
-            batch_results = await self._send_batch_messages(queue, queue_msgs)
+            batch_results = self._send_batch_messages_sync(queue, queue_msgs)
             results.extend(batch_results)
-        
+
+        return results
+
+    async def _send_tasks_async(self, messages: list):
+        """异步发送任务"""
+        if not messages:
+            return []
+
+        results = []
+
+        # 按队列分组消息，以便批量处理
+        queue_messages = {}
+        for msg in messages:
+            # 支持TaskMessage对象或字典
+            if isinstance(msg, dict):
+                msg = TaskMessage.from_dict(msg)
+            elif not isinstance(msg, TaskMessage):
+                raise ValueError(f"Invalid message type: {type(msg)}. Expected TaskMessage or dict")
+
+            # 验证消息
+            msg.validate()
+
+            # 确定实际的队列名（考虑优先级）
+            actual_queue = msg.queue
+            if msg.priority is not None:
+                # 将优先级拼接到队列名后面
+                actual_queue = f"{msg.queue}:{msg.priority}"
+                # 更新消息体中的queue字段，确保与实际发送的stream key一致
+                msg.queue = actual_queue
+
+            # 按队列分组
+            if actual_queue not in queue_messages:
+                queue_messages[actual_queue] = []
+            queue_messages[actual_queue].append(msg)
+
+        # 处理每个队列的消息
+        for queue, queue_msgs in queue_messages.items():
+            batch_results = await self._send_batch_messages_async(queue, queue_msgs)
+            results.extend(batch_results)
+
         return results
     
-    async def _send_batch_messages(self, queue: str, messages: list) -> list:
-        """批量发送single模式消息（内部方法）"""
+    def _send_batch_messages_sync(self, queue: str, messages: list) -> list:
+        """批量发送任务（同步）"""
         from ..utils.serializer import dumps_str
-        
+
         # 分离普通任务和延迟任务
         normal_messages = []
         delayed_messages = []
-        
+
         for msg in messages:
             msg_dict = msg.to_dict()
-            
+
             # 处理延迟任务
             if msg.delay and msg.delay > 0:
                 # 添加延迟执行标记
@@ -763,9 +779,48 @@ class Jettask(object):
                 delayed_messages.append((msg_dict, msg.delay))
             else:
                 normal_messages.append(msg_dict)
-        
+
         results = []
-        
+
+        # 发送普通任务（统一使用批量发送）
+        if normal_messages:
+            batch_results = self.ep._batch_send_event_sync(
+                self.ep.get_prefixed_queue_name(queue),
+                [{'data': dumps_str(msg)} for msg in normal_messages],
+                self.ep.get_redis_client(asyncio=False, binary=True).pipeline()
+            )
+            results.extend(batch_results)
+
+        # 发送延迟任务（需要同时添加到DELAYED_QUEUE）
+        if delayed_messages:
+            delayed_results = self._send_delayed_tasks_sync(queue, delayed_messages)
+            results.extend(delayed_results)
+
+        return results
+
+    async def _send_batch_messages_async(self, queue: str, messages: list) -> list:
+        """批量发送任务（异步）"""
+        from ..utils.serializer import dumps_str
+
+        # 分离普通任务和延迟任务
+        normal_messages = []
+        delayed_messages = []
+
+        for msg in messages:
+            msg_dict = msg.to_dict()
+
+            # 处理延迟任务
+            if msg.delay and msg.delay > 0:
+                # 添加延迟执行标记
+                current_time = time.time()
+                msg_dict['execute_at'] = current_time + msg.delay
+                msg_dict['is_delayed'] = 1
+                delayed_messages.append((msg_dict, msg.delay))
+            else:
+                normal_messages.append(msg_dict)
+
+        results = []
+
         # 发送普通任务（统一使用批量发送）
         if normal_messages:
             batch_results = await self.ep._batch_send_event(
@@ -774,89 +829,799 @@ class Jettask(object):
                 self.ep.get_redis_client(asyncio=True, binary=True).pipeline()
             )
             results.extend(batch_results)
-        
+
         # 发送延迟任务（需要同时添加到DELAYED_QUEUE）
         if delayed_messages:
-            delayed_results = await self._send_delayed_tasks(queue, delayed_messages)
+            delayed_results = await self._send_delayed_tasks_async(queue, delayed_messages)
             results.extend(delayed_results)
-        
+
         return results
     
-    async def _send_delayed_tasks(self, queue: str, delayed_messages: list) -> list:
-        """发送延迟任务到Stream并添加到延迟队列"""
-        from ..utils.serializer import dumps_str
-        
-        # 使用Lua脚本原子性地处理延迟任务
-        lua_script = """
-        local prefix = ARGV[1]
-        local results = {}
-        
-        -- 从ARGV[2]开始，每4个参数为一组任务信息
-        -- [stream_key, stream_data, execute_at, queue]
-        for i = 2, #ARGV, 4 do
-            local stream_key = ARGV[i]
-            local stream_data = ARGV[i+1]
-            local execute_at = tonumber(ARGV[i+2])
-            local queue_name = ARGV[i+3]
-            
-            -- 使用Hash存储所有队列的offset
-            local offsets_hash = prefix .. ':QUEUE_OFFSETS'
-            
-            -- 从stream_key中提取队列名
-            local queue_name = string.gsub(stream_key, '^' .. prefix .. ':QUEUE:', '')
-            
-            -- 使用HINCRBY原子递增offset
-            local current_offset = redis.call('HINCRBY', offsets_hash, queue_name, 1)
-            
-            -- 1. 添加消息到Stream（包含offset字段）
-            local stream_id = redis.call('XADD', stream_key, '*', 
-                'data', stream_data,
-                'offset', current_offset)
-            
-            -- 2. 添加到延迟队列ZSET
-            local delayed_queue_key = prefix .. ':DELAYED_QUEUE:' .. queue_name
-            redis.call('ZADD', delayed_queue_key, execute_at, stream_id)
-            
-            -- 3. 设置任务状态Hash
-            local task_key = prefix .. ':TASK:' .. stream_id
-            redis.call('HSET', task_key, 'status', 'delayed')
-            redis.call('EXPIRE', task_key, 3600)
-            
-            -- 保存stream_id到结果
-            table.insert(results, stream_id)
-        end
-        
-        return results
+    def _send_delayed_tasks_sync(self, queue: str, delayed_messages: list) -> list:
+        """发送延迟任务到Stream并添加到延迟队列（同步）
+
+        Args:
+            queue: 队列名，可能包含优先级后缀（如 "queue_name:6"）
+            delayed_messages: 延迟消息列表，每项为 (msg_dict, delay_seconds)
+
+        Note:
+            延迟队列和Stream现在完全对应（包括优先级后缀）。
+            Scanner会动态发现所有优先级队列并扫描对应的延迟队列。
         """
-        
+        from ..utils.serializer import dumps_str
+        from ..messaging.registry import QueueRegistry
+
+        # 注册队列（确保队列在注册表中，Scanner才能发现它）
+        registry = QueueRegistry(self.redis, self.async_redis, self.redis_prefix)
+
+        # 如果队列包含优先级后缀，注册为优先级队列
+        if ':' in queue and queue.rsplit(':', 1)[1].isdigit():
+            base_queue = queue.rsplit(':', 1)[0]
+            priority = int(queue.rsplit(':', 1)[1])
+            registry.register_queue_sync(base_queue)
+            registry.register_priority_queue_sync(base_queue, priority)
+            logger.debug(f"Registered priority queue: {queue} (base: {base_queue}, priority: {priority})")
+        else:
+            registry.register_queue_sync(queue)
+            logger.debug(f"Registered queue: {queue}")
+
         # 准备Lua脚本参数
         lua_args = [self.redis_prefix]
         prefixed_queue = self.ep.get_prefixed_queue_name(queue)
-        
-        for msg_dict, delay_seconds in delayed_messages:
+
+        for msg_dict, _ in delayed_messages:
             stream_data = dumps_str(msg_dict)
             execute_at = msg_dict['execute_at']
-            
+
+            # 延迟队列名和Stream名现在完全对应
             lua_args.extend([
-                prefixed_queue,
-                stream_data,
-                str(execute_at),
-                queue
+                prefixed_queue,        # Stream 键（包含优先级）
+                stream_data,           # 消息数据
+                str(execute_at)        # 执行时间
             ])
-        
+
+        # 执行Lua脚本
+        client = self.ep.get_redis_client(asyncio=False, binary=True)
+
+        # 注册Lua脚本（使用config模块中的脚本）
+        if not hasattr(self, '_delayed_task_script_sync'):
+            self._delayed_task_script_sync = client.register_script(LUA_SCRIPT_SEND_DELAYED_TASKS)
+
+        # 执行脚本
+        results = self._delayed_task_script_sync(keys=[], args=lua_args)
+
+        # 解码结果
+        decoded_results = [r.decode('utf-8') if isinstance(r, bytes) else r for r in results]
+        return decoded_results
+
+    async def _send_delayed_tasks_async(self, queue: str, delayed_messages: list) -> list:
+        """发送延迟任务到Stream并添加到延迟队列（异步）
+
+        Args:
+            queue: 队列名，可能包含优先级后缀（如 "queue_name:6"）
+            delayed_messages: 延迟消息列表，每项为 (msg_dict, delay_seconds)
+
+        Note:
+            延迟队列和Stream现在完全对应（包括优先级后缀）。
+            Scanner会动态发现所有优先级队列并扫描对应的延迟队列。
+        """
+        from ..utils.serializer import dumps_str
+        from ..messaging.registry import QueueRegistry
+
+        # 注册队列（确保队列在注册表中，Scanner才能发现它）
+        registry = QueueRegistry(self.redis, self.async_redis, self.redis_prefix)
+
+        # 如果队列包含优先级后缀，注册为优先级队列
+        if ':' in queue and queue.rsplit(':', 1)[1].isdigit():
+            base_queue = queue.rsplit(':', 1)[0]
+            priority = int(queue.rsplit(':', 1)[1])
+            await registry.register_queue(base_queue)
+            await registry.register_priority_queue(base_queue, priority)
+            logger.debug(f"Registered priority queue: {queue} (base: {base_queue}, priority: {priority})")
+        else:
+            await registry.register_queue(queue)
+            logger.debug(f"Registered queue: {queue}")
+
+        # 准备Lua脚本参数
+        lua_args = [self.redis_prefix]
+        prefixed_queue = self.ep.get_prefixed_queue_name(queue)
+
+        for msg_dict, _ in delayed_messages:
+            stream_data = dumps_str(msg_dict)
+            execute_at = msg_dict['execute_at']
+
+            # 延迟队列名和Stream名现在完全对应
+            lua_args.extend([
+                prefixed_queue,        # Stream 键（包含优先级）
+                stream_data,           # 消息数据
+                str(execute_at)        # 执行时间
+            ])
+
         # 执行Lua脚本
         client = self.ep.get_redis_client(asyncio=True, binary=True)
-        
-        # 注册Lua脚本
-        if not hasattr(self, '_delayed_task_script'):
-            self._delayed_task_script = client.register_script(lua_script)
-        
+
+        # 注册Lua脚本（使用config模块中的脚本）
+        if not hasattr(self, '_delayed_task_script_async'):
+            self._delayed_task_script_async = client.register_script(LUA_SCRIPT_SEND_DELAYED_TASKS)
+
         # 执行脚本
-        results = await self._delayed_task_script(keys=[], args=lua_args)
-        
+        results = await self._delayed_task_script_async(keys=[], args=lua_args)
+
         # 解码结果
-        return [r.decode('utf-8') if isinstance(r, bytes) else r for r in results]
-    
+        decoded_results = [r.decode('utf-8') if isinstance(r, bytes) else r for r in results]
+        return decoded_results
+
+    def _get_task_names_from_queue(self, queue: str, task_name: str = None) -> list:
+        """获取队列的任务名列表
+
+        Args:
+            queue: 队列名称（可能包含优先级后缀）
+            task_name: 可选的任务名，如果提供则直接返回 [task_name]
+
+        Returns:
+            任务名列表，如果 task_name 提供则返回 [task_name]，否则返回队列的所有任务名
+        """
+        if task_name is not None:
+            return [task_name]
+
+        # 确保 registry 有 redis_client
+        if self.registry.redis is None:
+            self.registry.redis = self.redis
+            self.registry.async_redis = self.async_redis
+
+        # 从 base_queue 中提取基础队列名（去掉优先级）
+        base_queue = queue.split(':')[0] if ':' in queue else queue
+        task_names = self.registry.get_task_names_by_queue_sync(base_queue)
+
+        return list(task_names) if task_names else []
+
+    def get_result(self, event_id: str, queue: str, task_name: str = None,
+                   delete: bool = False, asyncio: bool = False,
+                   delayed_deletion_ex: int = None, wait: bool = False,
+                   timeout: int = 300, poll_interval: float = 0.5):
+        """获取任务执行结果
+
+        在任务组架构下，每个任务都有独立的执行结果存储。
+        结果存储格式: {redis_prefix}:TASK:{event_id}:{group_name}
+
+        这个方法支持完全解耦的生产者-消费者模式，生产者只需要知道：
+        - event_id: 发送任务时返回的事件ID
+        - queue: 队列名称
+        - task_name: 任务名称（可选，不提供时会获取该队列所有任务的结果）
+
+        Args:
+            event_id: 任务事件ID（发送任务时返回的消息ID）
+            queue: 队列名称
+            task_name: 任务名称（可选）。如果不提供，会获取该队列所有任务的结果，返回列表
+            delete: 是否删除结果（默认False）
+            asyncio: 是否使用异步模式（默认False）
+            delayed_deletion_ex: 延迟删除时间（秒），设置后会在指定时间后自动删除
+            wait: 是否阻塞等待直到任务完成（默认False）
+            timeout: 等待超时时间（秒），默认300秒
+            poll_interval: 轮询间隔（秒），默认0.5秒
+
+        Returns:
+            当指定task_name时:
+                同步模式: 任务结果（字符串或字节），如果任务未完成返回None
+                异步模式: 返回协程，需要使用 await
+            当不指定task_name时:
+                返回列表，每个元素是字典: [{"task_name": "xxx", "result": ..., "status": ...}, ...]
+
+        Raises:
+            TaskTimeoutError: 等待超时
+            TaskExecutionError: 任务执行失败
+            TaskNotFoundError: 任务不存在
+
+        Examples:
+            # 获取单个任务结果
+            result = app.get_result("1234567890-0", "my_queue", task_name="my_task")
+
+            # 获取队列中所有任务的结果
+            results = app.get_result("1234567890-0", "my_queue")
+            # 返回: [{"task_name": "task1", "result": ..., "status": ...}, {"task_name": "task2", ...}]
+
+            # 异步获取所有任务结果
+            results = await app.get_result("1234567890-0", "my_queue", asyncio=True)
+        """
+        # 判断是否指定了 task_name，决定最终返回格式
+        return_single = task_name is not None
+
+        # 获取需要查询的任务名列表
+        task_names = self._get_task_names_from_queue(queue, task_name)
+
+        # 如果没有任务，直接返回空列表
+        if not task_names:
+            if asyncio:
+                async def _return_empty_list():
+                    return []
+                return _return_empty_list()
+            else:
+                return []
+
+        # 统一处理：遍历所有任务获取结果
+        if asyncio:
+            return self._get_results_async(event_id, queue, task_names, delete,
+                                          delayed_deletion_ex, wait, timeout, poll_interval, return_single)
+        else:
+            return self._get_results_sync(event_id, queue, task_names, delete,
+                                         delayed_deletion_ex, wait, timeout, poll_interval, return_single)
+
+    def get_queue_position(self, event_id: str, queue: str, task_name: str = None, asyncio: bool = False):
+        """获取任务在队列中的排队情况
+
+        通过 event_id 查询任务在队列中的排队位置，包括：
+        - 距离被读取还差多少任务
+        - 距离被消费还差多少任务
+
+        Args:
+            event_id: 任务事件ID（发送任务时返回的消息ID）
+            queue: 队列名称
+            task_name: 任务名称（可选）。如果不提供，会获取该队列所有任务的排队情况
+            asyncio: 是否使用异步模式（默认False）
+
+        Returns:
+            当指定task_name时:
+                返回字典: {
+                    "task_name": "xxx",
+                    "task_offset": 12,
+                    "read_offset": 14,
+                    "task_ack_offset": 10,
+                    "pending_read": 2,      # 距离被读取还差2个任务
+                    "pending_consume": -2   # 已经被消费了（负数表示已完成）
+                }
+            当不指定task_name时:
+                返回列表，每个元素是上述格式的字典
+
+        Note:
+            排名信息反映的是任务的发送顺序（offset），而不是执行顺序。
+            在并发执行的场景下，排名靠后的任务可能先执行完成，而排名靠前的任务可能还在执行中。
+
+            例如：
+            - 任务A (offset=10) 和任务B (offset=15) 同时被读取
+            - 如果任务B执行得快，可能会先完成
+            - 此时任务A的 pending_consume 可能仍为正数（还未消费确认）
+            - 而任务B的 pending_consume 已经变为负数（已完成）
+
+            因此：
+            - pending_read 表示有多少任务在你之前被发送到队列
+            - pending_consume 表示有多少任务在你之前被消费确认（不代表执行顺序）
+            - 负数的 pending_consume 只表示该任务已被确认，不表示所有前面的任务都已完成
+
+        Examples:
+            # 获取单个任务的排队情况
+            position = app.get_queue_position("1234567890-0", "my_queue", task_name="my_task")
+
+            # 获取队列中所有任务的排队情况
+            positions = app.get_queue_position("1234567890-0", "my_queue")
+
+            # 异步获取
+            position = await app.get_queue_position("1234567890-0", "my_queue", asyncio=True)
+        """
+        # 判断是否指定了 task_name，决定最终返回格式
+        return_single = task_name is not None
+
+        # 获取需要查询的任务名列表
+        task_names = self._get_task_names_from_queue(queue, task_name)
+
+        # 如果没有任务，直接返回空列表
+        if not task_names:
+            if asyncio:
+                async def _return_empty_list():
+                    return []
+                return _return_empty_list()
+            else:
+                return []
+
+        # 统一处理：遍历所有任务获取排队情况
+        if asyncio:
+            return self._get_queue_positions_async(event_id, queue, task_names, return_single)
+        else:
+            return self._get_queue_positions_sync(event_id, queue, task_names, return_single)
+
+    def _get_queue_positions_sync(self, event_id: str, queue: str, task_names: list, return_single: bool):
+        """同步获取任务排队情况"""
+        results = []
+
+        # 构建 stream key
+        prefixed_queue = f"{self.redis_prefix}:QUEUE:{queue}"
+
+        # 从 stream 中获取任务数据
+        try:
+            # XRANGE 获取指定 event_id 的消息
+            stream_data = self.binary_redis.xrange(prefixed_queue, min=event_id, max=event_id, count=1)
+
+            if not stream_data:
+                # 任务不存在于 stream 中
+                for task_name in task_names:
+                    results.append({
+                        "task_name": task_name,
+                        "error": "Task not found in stream"
+                    })
+                if return_single:
+                    return results[0] if results else None
+                return results
+
+            # 解析 stream 数据
+            message_id, message_data = stream_data[0]
+
+            # 解码字段
+            task_offset = None
+            for key, value in message_data.items():
+                if isinstance(key, bytes):
+                    key = key.decode('utf-8')
+                if key == 'offset':
+                    if isinstance(value, bytes):
+                        task_offset = int(value.decode('utf-8'))
+                    else:
+                        task_offset = int(value)
+                    break
+
+            if task_offset is None:
+                for task_name in task_names:
+                    results.append({
+                        "task_name": task_name,
+                        "error": "Offset not found in task data"
+                    })
+                if return_single:
+                    return results[0] if results else None
+                return results
+
+        except Exception as e:
+            for task_name in task_names:
+                results.append({
+                    "task_name": task_name,
+                    "error": f"Failed to read from stream: {str(e)}"
+                })
+            if return_single:
+                return results[0] if results else None
+            return results
+
+        # 获取 READ_OFFSETS 和 TASK_OFFSETS（使用 pipeline + HMGET 优化）
+        read_offsets_key = f"{self.redis_prefix}:READ_OFFSETS"
+        task_offsets_key = f"{self.redis_prefix}:TASK_OFFSETS"
+
+        # 提取基础队列名（去掉优先级）
+        base_queue = queue.split(':')[0] if ':' in queue else queue
+
+        # 构建需要查询的字段列表
+        offset_keys = [f"{base_queue}:{task_name}" for task_name in task_names]
+
+        try:
+            # 使用 pipeline 批量获取所有需要的字段
+            pipeline = self.redis.pipeline()
+            pipeline.hmget(read_offsets_key, offset_keys)
+            pipeline.hmget(task_offsets_key, offset_keys)
+            read_offsets_list, task_offsets_list = pipeline.execute()
+        except Exception as e:
+            for task_name in task_names:
+                results.append({
+                    "task_name": task_name,
+                    "error": f"Failed to read offsets: {str(e)}"
+                })
+            if return_single:
+                return results[0] if results else None
+            return results
+
+        # 对每个任务计算排队情况
+        for idx, task_name in enumerate(task_names):
+            # 获取 read_offset
+            read_offset = read_offsets_list[idx]
+            if read_offset is not None:
+                read_offset = int(read_offset)
+
+            # 获取 task_ack_offset
+            task_ack_offset = task_offsets_list[idx]
+            if task_ack_offset is not None:
+                task_ack_offset = int(task_ack_offset)
+
+            # 计算排队情况
+            # pending_read: 正数表示还差多少个任务才能被读取，0表示刚好被读取，负数表示已被读取
+            pending_read = (task_offset - read_offset) if read_offset is not None else None
+            # pending_consume: 正数表示还差多少个任务才能被消费，0表示刚好被消费，负数表示已被消费
+            pending_consume = (task_offset - task_ack_offset) if task_ack_offset is not None else None
+
+            results.append({
+                "task_name": task_name,
+                "task_offset": task_offset,
+                "read_offset": read_offset,
+                "task_ack_offset": task_ack_offset,
+                "pending_read": pending_read,
+                "pending_consume": pending_consume
+            })
+
+        # 根据 return_single 决定返回格式
+        if return_single:
+            return results[0] if results else None
+        return results
+
+    async def _get_queue_positions_async(self, event_id: str, queue: str, task_names: list, return_single: bool):
+        """异步获取任务排队情况"""
+        results = []
+
+        # 构建 stream key
+        prefixed_queue = f"{self.redis_prefix}:QUEUE:{queue}"
+
+        # 从 stream 中获取任务数据
+        try:
+            # XRANGE 获取指定 event_id 的消息
+            stream_data = await self.async_binary_redis.xrange(prefixed_queue, min=event_id, max=event_id, count=1)
+
+            if not stream_data:
+                # 任务不存在于 stream 中
+                for task_name in task_names:
+                    results.append({
+                        "task_name": task_name,
+                        "error": "Task not found in stream"
+                    })
+                if return_single:
+                    return results[0] if results else None
+                return results
+
+            # 解析 stream 数据
+            message_id, message_data = stream_data[0]
+
+            # 解码字段
+            task_offset = None
+            for key, value in message_data.items():
+                if isinstance(key, bytes):
+                    key = key.decode('utf-8')
+                if key == 'offset':
+                    if isinstance(value, bytes):
+                        task_offset = int(value.decode('utf-8'))
+                    else:
+                        task_offset = int(value)
+                    break
+
+            if task_offset is None:
+                for task_name in task_names:
+                    results.append({
+                        "task_name": task_name,
+                        "error": "Offset not found in task data"
+                    })
+                if return_single:
+                    return results[0] if results else None
+                return results
+
+        except Exception as e:
+            for task_name in task_names:
+                results.append({
+                    "task_name": task_name,
+                    "error": f"Failed to read from stream: {str(e)}"
+                })
+            if return_single:
+                return results[0] if results else None
+            return results
+
+        # 获取 READ_OFFSETS 和 TASK_OFFSETS（使用 pipeline + HMGET 优化）
+        read_offsets_key = f"{self.redis_prefix}:READ_OFFSETS"
+        task_offsets_key = f"{self.redis_prefix}:TASK_OFFSETS"
+
+        # 提取基础队列名（去掉优先级）
+        base_queue = queue.split(':')[0] if ':' in queue else queue
+
+        # 构建需要查询的字段列表
+        offset_keys = [f"{base_queue}:{task_name}" for task_name in task_names]
+
+        try:
+            # 使用 pipeline 批量获取所有需要的字段
+            pipeline = self.async_redis.pipeline()
+            pipeline.hmget(read_offsets_key, offset_keys)
+            pipeline.hmget(task_offsets_key, offset_keys)
+            read_offsets_list, task_offsets_list = await pipeline.execute()
+        except Exception as e:
+            for task_name in task_names:
+                results.append({
+                    "task_name": task_name,
+                    "error": f"Failed to read offsets: {str(e)}"
+                })
+            if return_single:
+                return results[0] if results else None
+            return results
+
+        # 对每个任务计算排队情况
+        for idx, task_name in enumerate(task_names):
+            # 获取 read_offset
+            read_offset = read_offsets_list[idx]
+            if read_offset is not None:
+                read_offset = int(read_offset)
+
+            # 获取 task_ack_offset
+            task_ack_offset = task_offsets_list[idx]
+            if task_ack_offset is not None:
+                task_ack_offset = int(task_ack_offset)
+
+            # 计算排队情况
+            # pending_read: 正数表示还差多少个任务才能被读取，0表示刚好被读取，负数表示已被读取
+            pending_read = (task_offset - read_offset) if read_offset is not None else None
+            # pending_consume: 正数表示还差多少个任务才能被消费，0表示刚好被消费，负数表示已被消费
+            pending_consume = (task_offset - task_ack_offset) if task_ack_offset is not None else None
+
+            results.append({
+                "task_name": task_name,
+                "task_offset": task_offset,
+                "read_offset": read_offset,
+                "task_ack_offset": task_ack_offset,
+                "pending_read": pending_read,
+                "pending_consume": pending_consume
+            })
+
+        # 根据 return_single 决定返回格式
+        if return_single:
+            return results[0] if results else None
+        return results
+
+    def _build_task_key(self, task_name: str, queue: str, event_id: str):
+        """构建任务的 key 信息
+
+        Returns:
+            tuple: (group_name, full_key)
+        """
+        prefixed_queue = f"{self.redis_prefix}:QUEUE:{queue}"
+        group_name = f"{prefixed_queue}:{task_name}"
+        status_key = f"{event_id}:{group_name}"
+        full_key = f"{self.redis_prefix}:TASK:{status_key}"
+        return group_name, full_key
+
+    @staticmethod
+    def _decode_bytes(value):
+        """解码字节为字符串"""
+        if value and isinstance(value, bytes):
+            return value.decode('utf-8')
+        return value
+
+    @staticmethod
+    def _is_task_completed(status):
+        """检查任务是否已完成（成功）"""
+        return status in [TaskStatus.COMPLETED.value, TaskStatus.SUCCESS.value]
+
+    @staticmethod
+    def _is_task_failed(status):
+        """检查任务是否失败"""
+        return status in [TaskStatus.ERROR.value, TaskStatus.FAILED.value, "ERROR", "FAILED", "error", "failed"]
+
+    def _get_results_sync(self, event_id: str, queue: str, task_names: list,
+                         delete: bool, delayed_deletion_ex: int, wait: bool,
+                         timeout: int, poll_interval: float, return_single: bool):
+        """同步获取任务结果（支持单个或批量）"""
+        results = []
+
+        for task_name in task_names:
+            try:
+                _, full_key = self._build_task_key(task_name, queue, event_id)
+
+                # 统一调用 _get_result_sync，通过 wait 参数控制行为
+                task_info = self._get_result_sync(full_key, event_id, delete, delayed_deletion_ex,
+                                                  wait, timeout, poll_interval)
+
+                # 如果任务不存在
+                if not task_info:
+                    results.append({
+                        "task_name": task_name,
+                        "status": None,
+                        "result": None
+                    })
+                else:
+                    # 添加 task_name 到结果中
+                    task_info["task_name"] = task_name
+                    results.append(task_info)
+
+            except Exception as e:
+                results.append({
+                    "task_name": task_name,
+                    "status": "ERROR",
+                    "result": None,
+                    "error_msg": str(e)
+                })
+
+        # 根据 return_single 决定返回格式
+        if return_single:
+            return results[0] if results else None
+        return results
+
+    async def _get_results_async(self, event_id: str, queue: str, task_names: list,
+                                 delete: bool, delayed_deletion_ex: int, wait: bool,
+                                 timeout: int, poll_interval: float, return_single: bool):
+        """异步获取任务结果（支持单个或批量）"""
+        results = []
+
+        for task_name in task_names:
+            try:
+                _, full_key = self._build_task_key(task_name, queue, event_id)
+
+                # 统一调用 _get_result_async，通过 wait 参数控制行为
+                task_info = await self._get_result_async(full_key, event_id, delete, delayed_deletion_ex,
+                                                         wait, timeout, poll_interval)
+
+                # 如果任务不存在
+                if not task_info:
+                    results.append({
+                        "task_name": task_name,
+                        "status": TaskStatus.PENDING.value,
+                        "result": None
+                    })
+                else:
+                    # 添加 task_name 到结果中
+                    task_info["task_name"] = task_name
+                    results.append(task_info)
+
+            except Exception as e:
+                results.append({
+                    "task_name": task_name,
+                    "status": "ERROR",
+                    "result": None,
+                    "error_msg": str(e)
+                })
+
+        # 根据 return_single 决定返回格式
+        if return_single:
+            return results[0] if results else None
+        return results
+
+    def _get_result_sync(self, full_key: str, event_id: str, delete: bool, delayed_deletion_ex: int,
+                         wait: bool = False, timeout: int = 300, poll_interval: float = 0.5):
+        """同步获取任务结果（支持等待模式）"""
+        from ..exceptions import TaskTimeoutError, TaskExecutionError, TaskNotFoundError
+
+        # 使用二进制客户端，不自动解码（因为 result 是 msgpack 序列化的）
+        client = self.binary_redis
+        start_time = time.time()
+
+        while True:
+            # 获取整个 hash 的所有字段
+            task_data = client.hgetall(full_key)
+
+            if not task_data:
+                if wait:
+                    raise TaskNotFoundError(f"Task {event_id} not found")
+                return None
+
+            # 解码字节字段
+            decoded_data = {}
+            for key, value in task_data.items():
+                # 解码 key
+                if isinstance(key, bytes):
+                    key = key.decode('utf-8')
+
+                # 跳过内部标记字段
+                if key.startswith('__'):
+                    continue
+
+                # 解码 value - 只有 result 字段需要 loads_str
+                if isinstance(value, bytes):
+                    if key == 'result':
+                        try:
+                            decoded_data[key] = loads_str(value)
+                        except Exception:
+                            decoded_data[key] = value
+                    else:
+                        # 其他字段尝试 UTF-8 解码
+                        try:
+                            decoded_data[key] = value.decode('utf-8')
+                        except Exception:
+                            decoded_data[key] = value
+                else:
+                    decoded_data[key] = value
+
+            # 如果不需要等待，处理删除逻辑后直接返回
+            if not wait:
+                if delayed_deletion_ex is not None:
+                    client.expire(full_key, delayed_deletion_ex)
+                elif delete:
+                    if self.task_center and self.task_center.is_enabled:
+                        client.hset(full_key, "__pending_delete", "1")
+                    else:
+                        client.delete(full_key)
+                return decoded_data
+
+            # 需要等待：检查任务状态
+            status = decoded_data.get('status')
+
+            # 检查任务是否完成
+            if self._is_task_completed(status):
+                # 任务成功完成，处理删除逻辑后返回
+                if delayed_deletion_ex is not None:
+                    client.expire(full_key, delayed_deletion_ex)
+                elif delete:
+                    if self.task_center and self.task_center.is_enabled:
+                        client.hset(full_key, "__pending_delete", "1")
+                    else:
+                        client.delete(full_key)
+                return decoded_data
+
+            elif self._is_task_failed(status):
+                # 任务失败，抛出异常
+                error_msg = decoded_data.get('error_msg', 'Task execution failed')
+                raise TaskExecutionError(event_id, error_msg)
+
+            # 检查超时
+            if time.time() - start_time > timeout:
+                raise TaskTimeoutError(f"Task {event_id} timed out after {timeout} seconds")
+
+            # 任务仍在执行中，等待后重试
+            time.sleep(poll_interval)
+
+    async def _get_result_async(self, full_key: str, event_id: str, delete: bool, delayed_deletion_ex: int,
+                                wait: bool = False, timeout: int = 300, poll_interval: float = 0.5):
+        """异步获取任务结果（支持等待模式）"""
+
+        # 使用二进制客户端，不自动解码（因为 result 是 msgpack 序列化的）
+        client = self.async_binary_redis
+        start_time = time.time()
+
+        while True:
+            # 获取整个 hash 的所有字段
+            task_data = await client.hgetall(full_key)
+
+            if not task_data:
+                if wait:
+                    raise TaskNotFoundError(f"Task {event_id} not found")
+                return None
+
+            # 解码字节字段
+            decoded_data = {}
+            for key, value in task_data.items():
+                # 解码 key
+                if isinstance(key, bytes):
+                    key = key.decode('utf-8')
+
+                # 跳过内部标记字段
+                if key.startswith('__'):
+                    continue
+
+                # 解码 value - 只有 result 字段需要 loads_str
+                if isinstance(value, bytes):
+                    if key == 'result':
+                        try:
+                            decoded_data[key] = loads_str(value)
+                        except Exception:
+                            decoded_data[key] = value
+                    else:
+                        # 其他字段尝试 UTF-8 解码
+                        try:
+                            decoded_data[key] = value.decode('utf-8')
+                        except Exception:
+                            decoded_data[key] = value
+                else:
+                    decoded_data[key] = value
+
+            # 如果不需要等待，处理删除逻辑后直接返回
+            if not wait:
+                if delayed_deletion_ex is not None:
+                    await client.expire(full_key, delayed_deletion_ex)
+                elif delete:
+                    if self.task_center and self.task_center.is_enabled:
+                        await client.hset(full_key, "__pending_delete", "1")
+                    else:
+                        await client.delete(full_key)
+                return decoded_data
+
+            # 需要等待：检查任务状态
+            status = decoded_data.get('status')
+
+            # 检查任务是否完成
+            if self._is_task_completed(status):
+                # 任务成功完成，处理删除逻辑后返回
+                if delayed_deletion_ex is not None:
+                    await client.expire(full_key, delayed_deletion_ex)
+                elif delete:
+                    if self.task_center and self.task_center.is_enabled:
+                        await client.hset(full_key, "__pending_delete", "1")
+                    else:
+                        await client.delete(full_key)
+                return decoded_data
+
+            elif self._is_task_failed(status):
+                # 任务失败，抛出异常
+                error_msg = decoded_data.get('error_msg', 'Task execution failed')
+                raise TaskExecutionError(event_id, error_msg)
+
+            # 检查超时
+            if time.time() - start_time > timeout:
+                raise TaskTimeoutError(f"Task {event_id} timed out after {timeout} seconds")
+
+            # 等待后重试
+            await asyncio.sleep(poll_interval)
+
     def register_router(self, router, prefix: str = None):
         """
         注册任务路由器
@@ -879,7 +1644,7 @@ class Jettask(object):
             app = Jettask(redis_url="redis://localhost:6379/0")
             app.register_router(email_router)
         """
-        from ..router import TaskRouter
+        from ..task.router import TaskRouter
         
         if not isinstance(router, TaskRouter):
             raise TypeError("router must be a TaskRouter instance")
@@ -899,7 +1664,7 @@ class Jettask(object):
             
             # 注册任务
             task = self._task_from_fun(func, name, None, queue, **task_config)
-            logger.info(f"Registered task: {name} (queue: {queue or self.redis_prefix})")
+            logger.debug(f"Registered task: {name} (queue: {queue or self.redis_prefix})")
         
         return self
 
@@ -942,25 +1707,159 @@ class Jettask(object):
             error_msg += f"3. 或者将这些任务的队列从监听列表中移除\n"
             raise ValueError(error_msg)
     
-    def _start(
+
+    def _start_with_heartbeat_thread(
         self,
-        execute_type: str = "asyncio",
+        execute_type: str = "multi_asyncio",
         queues: List[str] = None,
         concurrency: int = 1,
         prefetch_multiplier: int = 1,
-        **kwargs
     ):
+        """在主进程中启动执行器和心跳线程"""
+        from jettask.worker.lifecycle import HeartbeatThreadManager
+
+        # 1. 初始化 Worker ID - 复用 EventPool 中的 HeartbeatConsumerStrategy
+        # 确保 EventPool 已初始化
+        if not self.ep:
+            raise RuntimeError("EventPool not initialized")
+
+        # 使用 EventPool 中已创建的 HeartbeatConsumerStrategy
+        if not self.consumer_manager or not self.consumer_manager._heartbeat_strategy:
+            raise RuntimeError("ConsumerManager or HeartbeatConsumerStrategy not initialized")
+
+        strategy = self.consumer_manager._heartbeat_strategy
+        strategy._ensure_consumer_id()
+        worker_id = strategy.consumer_id
+
+        logger.info(f"Starting worker {worker_id} in main process (PID: {os.getpid()})")
+
+        # 2. 启动心跳线程（在主进程中）
+        heartbeat = HeartbeatThreadManager(
+            redis_client=self.redis,
+            worker_key=f"{self.redis_prefix}:WORKER:{worker_id}",
+            worker_id=worker_id,
+            redis_prefix=self.redis_prefix,
+            interval=5.0
+        )
+        heartbeat.start()
+
+        # 保存引用以便清理
+        self._heartbeat_manager = heartbeat
+        self._executor_processes = []
+
+        try:
+            # 3. 启动多进程执行器（直接调用 _start）
+            self._start(
+                execute_type=execute_type,
+                queues=queues,
+                concurrency=concurrency,
+                prefetch_multiplier=prefetch_multiplier,
+            )
+        except KeyboardInterrupt:
+            logger.info("Received interrupt signal")
+        finally:
+            # 4. 清理
+            logger.info("Shutting down worker...")
+
+            # 停止心跳线程
+            if hasattr(self, '_heartbeat_manager'):
+                logger.debug("Stopping heartbeat thread...")
+                self._heartbeat_manager.stop(timeout=2.0)
+
+            # 清理资源
+            self.cleanup()
+
+    # ==================== 新的子方法：重构后的 Worker 启动逻辑 ====================
+
+    def _generate_worker_id_lightweight(self) -> tuple:
+        """
+        轻量级生成 Worker ID（不初始化 EventPool）
+
+        这个方法只生成 worker_id，不会初始化 EventPool、ConsumerManager 等重量级组件。
+        用于主进程在 fork 子进程前生成 worker_id。
+
+        Returns:
+            (worker_id, worker_key) 元组
+        """
+        from jettask.worker.manager import WorkerNaming
+        import asyncio
+        import socket
+
+        # 生成主机名前缀（与 HeartbeatConsumerStrategy 相同的逻辑）
+        try:
+            hostname = socket.gethostname()
+            ip = socket.gethostbyname(hostname)
+            prefix = hostname if hostname != 'localhost' else ip
+        except:
+            prefix = os.environ.get('HOSTNAME', 'unknown')
+
+        # 创建轻量级的 worker naming
+        naming = WorkerNaming()
+
+        # 尝试复用离线的 worker ID
+        reusable_id = None
+        try:
+            loop = asyncio.get_event_loop()
+            if loop.is_running():
+                logger.warning("Event loop is running, cannot reuse offline worker ID synchronously")
+            else:
+                # 直接使用 WorkerNaming 查找可复用的 ID
+                reusable_id = loop.run_until_complete(
+                    naming.find_reusable_worker_id(
+                        prefix=prefix,
+                        worker_state=self.worker_state
+                    )
+                )
+        except RuntimeError:
+            # 没有事件循环，创建新的
+            loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(loop)
+            try:
+                reusable_id = loop.run_until_complete(
+                    naming.find_reusable_worker_id(
+                        prefix=prefix,
+                        worker_state=self.worker_state
+                    )
+                )
+            finally:
+                loop.close()
+
+        # 生成或复用 worker_id
+        if reusable_id:
+            logger.info(f"[PID {os.getpid()}] Reusing offline worker ID: {reusable_id}")
+            worker_id = reusable_id
+        else:
+            worker_id = naming.generate_worker_id(prefix)
+            logger.info(f"[PID {os.getpid()}] Generated new worker ID: {worker_id}")
+
+        worker_key = f"{self.redis_prefix}:WORKER:{worker_id}"
+
+        # 保存到实例
+        self.worker_id = worker_id
+        self.worker_key = worker_key
+
+        return worker_id, worker_key
+
+    def _initialize_worker(self, queues: List[str] = None) -> tuple:
+        """
+        初始化 Worker（完整版本，包含 EventPool 初始化）
+
+        注意：这个方法会初始化 EventPool 和 ConsumerManager，会创建事件循环等状态。
+        在多进程模式下，这些状态会被 fork 到子进程，需要在子进程中清理。
+
+        Returns:
+            (worker_id, worker_key) 元组
+        """
         # 设置默认队列
         if not queues:
             queues = [self.redis_prefix]
-        
+
+        # 初始化 EventPool
         self.ep.queues = queues
         self.ep.init_routing()
         self._mount_module()
-        # 验证任务兼容性 
-        self._validate_tasks_for_executor(execute_type, queues)
-        
-        # 收集每个队列上的所有任务（用于广播支持）
+
+        # 收集任务列表（按队列分组）
         self._tasks_by_queue = {}
         for task_name, task in self._tasks.items():
             task_queue = task.queue or self.redis_prefix
@@ -968,72 +1867,276 @@ class Jettask(object):
                 if task_queue not in self._tasks_by_queue:
                     self._tasks_by_queue[task_queue] = []
                 self._tasks_by_queue[task_queue].append(task_name)
-                logger.debug(f"Task {task_name} listening on queue {task_queue}")
-        
-        event_queue = deque()
-        
-        # 消费者组会在listening_event方法内部自动创建
-        
-        # 根据执行器类型创建对应的执行器
-        if execute_type == "asyncio":
-            # 对于asyncio执行器，使用asyncio.Queue
-            async_event_queue = asyncio.Queue()
-            
-            async def run_asyncio_executor():
-                # 启动异步事件监听
-                asyncio.create_task(self.ep.listening_event(async_event_queue, prefetch_multiplier))
-                # 创建并运行执行器
-                executor = AsyncioExecutor(async_event_queue, self, concurrency)
-                await executor.loop()
-            
-            try:
-                loop = asyncio.get_event_loop()
-                if loop.is_running():
-                    # 如果事件循环已经在运行，创建一个新的
-                    loop = asyncio.new_event_loop()
-                    asyncio.set_event_loop(loop)
-            except RuntimeError:
-                # 如果当前线程没有事件循环，创建一个新的
-                loop = asyncio.new_event_loop()
-                asyncio.set_event_loop(loop)
-            
-            try:
-                loop.run_until_complete(run_asyncio_executor())
-            except RuntimeError as e:
-                if "Event loop stopped" in str(e):
-                    logger.info("Event loop stopped, shutting down gracefully")
+                logger.debug(f"Task {task_name} -> queue {task_queue}")
+
+        # 创建 Worker ID - 复用 EventPool 中的 HeartbeatConsumerStrategy
+        # 确保 EventPool 已初始化
+        if not self.ep:
+            raise RuntimeError("EventPool not initialized")
+
+        # 使用 EventPool 中已创建的 HeartbeatConsumerStrategy
+        if not self.consumer_manager or not self.consumer_manager._heartbeat_strategy:
+            raise RuntimeError("ConsumerManager or HeartbeatConsumerStrategy not initialized")
+
+        strategy = self.consumer_manager._heartbeat_strategy
+
+        # 尝试复用离线的 worker ID
+        import asyncio
+        try:
+            loop = asyncio.get_event_loop()
+            if loop.is_running():
+                # 如果事件循环正在运行，无法同步调用异步方法
+                # 直接生成新的 ID（这种情况通常不会发生）
+                logger.warning("Event loop is running, cannot reuse offline worker ID synchronously")
+                strategy._ensure_consumer_id()
+            else:
+                # 事件循环未运行，可以使用 run_until_complete
+                reusable_id = loop.run_until_complete(self._find_reusable_worker_id_async(strategy))
+                if reusable_id:
+                    logger.info(f"[PID {os.getpid()}] Reusing offline worker ID: {reusable_id}")
+                    strategy.consumer_id = reusable_id
+                    strategy._worker_key = f'{self.redis_prefix}:WORKER:{reusable_id}'
                 else:
-                    raise
-        elif execute_type == "multi_asyncio":
-            # multi_asyncio在每个子进程中会启动自己的监听器
-            executor = MultiAsyncioExecutor(event_queue, self, concurrency)
-            executor.prefetch_multiplier = prefetch_multiplier
-            
-            # 设置信号处理器以正确响应Ctrl+C
-            def multi_asyncio_signal_handler(signum, _frame):
-                logger.info(f"Multi-asyncio mode received signal {signum}")
-                executor._main_received_signal = True
-                executor.shutdown_event.set()
-                # 强制退出主循环
-                raise KeyboardInterrupt()
-            
-            signal.signal(signal.SIGINT, multi_asyncio_signal_handler)
-            signal.signal(signal.SIGTERM, multi_asyncio_signal_handler)
-            
+                    strategy._ensure_consumer_id()
+        except RuntimeError:
+            # 没有事件循环，创建新的
+            loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(loop)
             try:
-                executor.loop()
-            except KeyboardInterrupt:
-                logger.info("Multi-asyncio mode interrupted")
+                reusable_id = loop.run_until_complete(self._find_reusable_worker_id_async(strategy))
+                if reusable_id:
+                    logger.info(f"[PID {os.getpid()}] Reusing offline worker ID: {reusable_id}")
+                    strategy.consumer_id = reusable_id
+                    strategy._worker_key = f'{self.redis_prefix}:WORKER:{reusable_id}'
+                else:
+                    strategy._ensure_consumer_id()
             finally:
+                loop.close()
+
+        worker_id = strategy.consumer_id
+        worker_key = f"{self.redis_prefix}:WORKER:{worker_id}"
+
+        # 保存 worker_id 到实例，供子进程使用
+        self.worker_id = worker_id
+        self.worker_key = worker_key
+
+        logger.info(f"Worker initialized: {worker_id} (PID: {os.getpid()})")
+        return worker_id, worker_key
+
+    async def _find_reusable_worker_id_async(self, strategy) -> str:
+        """
+        异步查找可复用的离线 worker ID
+
+        Args:
+            strategy: HeartbeatConsumerStrategy 实例
+
+        Returns:
+            可复用的 worker ID，如果没有则返回 None
+        """
+        from jettask.worker.manager import WorkerNaming
+        naming = WorkerNaming()
+        return await naming.find_reusable_worker_id(
+            prefix=strategy.hostname_prefix,
+            worker_state=self.worker_state
+        )
+
+    def _start_heartbeat_thread_v2(self, worker_id: str, worker_key: str, queues: List[str] = None):
+        """
+        启动心跳线程
+
+        Args:
+            worker_id: Worker ID
+            worker_key: Worker Redis key
+            queues: Worker 负责的队列列表（用于消息恢复）
+
+        Returns:
+            HeartbeatThreadManager 实例
+        """
+        from jettask.worker.lifecycle import HeartbeatThreadManager
+
+        heartbeat = HeartbeatThreadManager(
+            redis_client=self.redis,
+            worker_key=worker_key,
+            worker_id=worker_id,
+            redis_prefix=self.redis_prefix,
+            interval=5.0
+        )
+
+        # 在启动心跳线程前设置 queues（确保第一次心跳就能写入 Redis）
+        if queues:
+            for queue in queues:
+                heartbeat.queues.add(queue)
+            logger.debug(f"Configured queues for heartbeat: {queues}")
+
+        heartbeat.start()
+        logger.info(f"Heartbeat thread started for worker {worker_id}")
+        return heartbeat
+
+    def _create_executor(self, concurrency: int):
+        """
+        创建进程编排器实例
+
+        Returns:
+            ProcessOrchestrator 实例
+        """
+        # 创建 ProcessOrchestrator（多进程管理器）
+        orchestrator = ProcessOrchestrator(self, concurrency)
+
+        # 保存 orchestrator 引用
+        self._current_executor = orchestrator
+
+        # 设置信号处理器
+        def signal_handler(signum, frame):
+            logger.info(f"Main process received signal {signum}, initiating shutdown...")
+            self._should_exit = True
+            orchestrator.shutdown_event.set()
+            raise KeyboardInterrupt()
+
+        signal.signal(signal.SIGINT, signal_handler)
+        signal.signal(signal.SIGTERM, signal_handler)
+
+        return orchestrator
+
+    def _cleanup_worker_v3(self, heartbeat_managers: list, executor, worker_ids: list):
+        """
+        清理 Worker 资源（新版本，支持多个心跳线程）
+        """
+        logger.info("Shutting down workers...")
+
+        # 1. 停止所有心跳线程
+        if heartbeat_managers:
+            logger.info(f"Stopping {len(heartbeat_managers)} heartbeat threads...")
+            for i, heartbeat in enumerate(heartbeat_managers):
+                try:
+                    worker_id = worker_ids[i][0] if i < len(worker_ids) else f"worker_{i}"
+                    logger.debug(f"Stopping heartbeat thread for {worker_id}...")
+                    heartbeat.stop(timeout=2.0)
+                except Exception as e:
+                    logger.error(f"Error stopping heartbeat #{i}: {e}", exc_info=True)
+
+        # 2. 关闭执行器
+        if executor:
+            try:
+                logger.debug("Shutting down executor...")
                 executor.shutdown()
-        else:
-            raise ValueError(f"不支持的执行器类型：{execute_type}，仅支持 'asyncio' 和 'multi_asyncio'")
+            except Exception as e:
+                logger.error(f"Error shutting down executor: {e}", exc_info=True)
+
+        # 3. 关闭 Redis 连接
+        try:
+            logger.debug("Closing Redis connections...")
+            self._cleanup_redis_connections_v2()
+        except Exception as e:
+            logger.error(f"Error closing Redis connections: {e}", exc_info=True)
+
+        # 4. 调用通用清理
+        try:
+            self.cleanup()
+        except Exception as e:
+            logger.error(f"Error in cleanup: {e}", exc_info=True)
+
+        logger.info(f"All {len(worker_ids)} workers shutdown complete")
+
+    def _cleanup_redis_connections_v2(self):
+        """清理 Redis 连接（异步包装）"""
+        async def async_cleanup_redis():
+            """异步关闭 Redis 连接"""
+            try:
+                logger.debug("Closing async Redis connections...")
+
+                # 关闭 EventPool 的连接
+                if hasattr(self.ep, 'async_redis_client'):
+                    await self.ep.async_redis_client.aclose()
+
+                if hasattr(self.ep, 'async_binary_redis_client'):
+                    await self.ep.async_binary_redis_client.aclose()
+
+                # 关闭 app 级别的连接
+                if hasattr(self, '_async_redis'):
+                    await self._async_redis.aclose()
+
+                logger.debug("Async Redis connections closed")
+            except Exception as e:
+                logger.error(f"Error closing async Redis: {e}", exc_info=True)
+
+        # 在新的事件循环中执行异步清理
+        try:
+            import asyncio
+            cleanup_loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(cleanup_loop)
+            cleanup_loop.run_until_complete(async_cleanup_redis())
+            cleanup_loop.close()
+        except Exception as e:
+            logger.error(f"Error in Redis cleanup: {e}", exc_info=True)
+
+
+    def _start(self, queues: List[str] = None, concurrency: int = 1, prefetch_multiplier: int = 1):
+        """
+        启动 Worker 主逻辑（多进程模式，主进程调用）
+
+        新架构流程：
+        1. 为每个子进程生成独立的 Worker ID
+        2. 在主进程为每个子进程启动独立的心跳线程
+        3. 创建执行器
+        4. Fork 并启动子进程，传递对应的 worker_id
+        5. 等待退出信号（阻塞）
+        6. 清理资源
+
+        Args:
+            queues: 监听的队列列表
+            concurrency: 并发执行器进程数（子进程数量）
+            prefetch_multiplier: 预取倍数
+        """
+        heartbeat_managers = []
+        executor = None
+        worker_ids = []
+
+        try:
+            # 1. 为每个子进程生成独立的 Worker ID 并启动心跳线程
+            logger.info(f"Generating {concurrency} worker IDs and starting heartbeat threads...")
+            from jettask.worker.lifecycle import HeartbeatThreadManager
+
+            for i in range(concurrency):
+                # 使用新方法：生成 worker_id 并启动心跳，等待首次心跳成功
+                heartbeat = HeartbeatThreadManager.create_and_start(
+                    redis_client=self.redis,
+                    redis_prefix=self.redis_prefix,
+                    queues=queues,
+                    interval=5.0,
+                    worker_state=self.worker_state
+                )
+                # 从心跳管理器对象中获取 worker_id 和 worker_key
+                worker_ids.append((heartbeat.worker_id, heartbeat.worker_key))
+                heartbeat_managers.append(heartbeat)
+                logger.info(f"  Process #{i}: worker_id={heartbeat.worker_id} (heartbeat started)")
+
+            # 2. 创建执行器
+            executor = self._create_executor(concurrency)
+
+            # 3. 启动 ProcessOrchestrator（阻塞调用，会fork多个子进程并运行）
+            # 传递 worker_ids 列表，每个子进程使用对应的 worker_id
+            logger.info(f"Starting {concurrency} executor processes...")
+
+            executor.start(
+                queues=queues,
+                prefetch_multiplier=prefetch_multiplier,
+                worker_ids=worker_ids  # 传递 worker_ids 列表
+            )
+
+        except KeyboardInterrupt:
+            logger.info("Worker interrupted by keyboard")
+        except Exception as e:
+            logger.error(f"Error in worker main loop: {e}", exc_info=True)
+        finally:
+            # 5. 清理资源
+            self._cleanup_worker_v3(heartbeat_managers, executor, worker_ids)
+
+    # ==================== 旧方法（待废弃） ====================
 
     def _run_subprocess(self, *args, **kwargs):
-        # logger.info("Started Worker Process")
-        process = multiprocessing.Process(target=self._start, args=args, kwargs=kwargs)
-        process.start()
-        return process
+        """已废弃：不再使用子进程"""
+        _ = (args, kwargs)  # 避免未使用警告
+        raise DeprecationWarning("_run_subprocess is deprecated, use _start_with_heartbeat_thread instead")
 
     def start(
         self,
@@ -1043,48 +2146,28 @@ class Jettask(object):
         prefetch_multiplier: int = 1,
         reload: bool = False,
     ):
+        """启动 Worker（仅支持 multi_asyncio）"""
+        _ = (reload, execute_type)  # 参数已废弃，避免未使用警告
+
         # 标记worker已启动
         self._worker_started = True
-        
+
         # 如果配置了任务中心且配置尚未加载，从任务中心获取配置
         if self.task_center and self.task_center.is_enabled and not self._task_center_config:
             self._load_config_from_task_center()
-        
+
         # 注册清理处理器（只在启动worker时注册）
         self._setup_cleanup_handlers()
-        
-        if execute_type == "multi_asyncio" and self.consumer_strategy == "pod":
+
+        if self.consumer_strategy == "pod":
             raise ValueError("multi_asyncio模式下无法使用pod策略")
-        self.process = self._run_subprocess(
-            execute_type=execute_type,
+
+        # 使用重构后的 _start() 方法
+        self._start(
             queues=queues,
             concurrency=concurrency,
             prefetch_multiplier=prefetch_multiplier,
         )
-        if reload:
-            event_handler = FileChangeHandler(
-                self,
-                execute_type=execute_type,
-                queues=queues,
-                concurrency=concurrency,
-                prefetch_multiplier=prefetch_multiplier,
-            )
-            observer = Observer()
-            observer.schedule(event_handler, ".", recursive=True)
-            observer.start()
-        # 使用事件来等待，而不是无限循环
-        try:
-            while not self._should_exit:
-                time.sleep(0.1)  # 短暂睡眠，快速响应退出信号
-        except KeyboardInterrupt:
-            logger.info("Received keyboard interrupt")
-            self.cleanup()
-            if self.process and self.process.is_alive():
-                self.process.terminate()
-                self.process.join(timeout=2)
-                if self.process.is_alive():
-                    logger.warning("Process did not terminate, killing...")
-                    self.process.kill()
 
 
     def get_task_info(self, event_id: str, asyncio: bool = False):
@@ -1097,16 +2180,20 @@ class Jettask(object):
             return client.hgetall(key)
     
     def get_task_status(self, event_id: str, asyncio: bool = False):
-        """获取任务状态（从TASK:hash的status字段）"""
+        """获取任务状态（从TASK:hash的status字段）
+
+        注意：这个方法使用简化的 key 格式 TASK:{event_id}
+        如果需要获取带 group_name 的任务状态，请使用 _get_task_status_sync 或 _get_task_status_async
+        """
         if asyncio:
-            return self._get_task_status_async(event_id)
+            return self._get_task_status_simple_async(event_id)
         else:
             client = self.redis
             key = f"{self.redis_prefix}:TASK:{event_id}"
             return client.hget(key, "status")
 
-    async def _get_task_status_async(self, event_id: str):
-        """异步获取任务状态"""
+    async def _get_task_status_simple_async(self, event_id: str):
+        """异步获取任务状态（简化版本，不需要 task_name 和 queue）"""
         key = f"{self.redis_prefix}:TASK:{event_id}"
         return await self.async_redis.hget(key, "status")
 
@@ -1152,199 +2239,18 @@ class Jettask(object):
     def get_redis_client(self, asyncio: bool = False):
         return self.async_redis if asyncio else self.redis
 
-    def set_data(
-        self, event_id: str, result: str, ex: int = 3600, asyncio: bool = False
-    ):
-        """设置任务结果（写入TASK:hash的result字段）"""
-        client = self.get_redis_client(asyncio)
-        key = f"{self.redis_prefix}:TASK:{event_id}"
-        if asyncio:
-            return self._set_data_async(key, result, ex)
-        else:
-            client.hset(key, "result", result)
-            return client.expire(key, ex)
-    
-    async def _set_data_async(self, key: str, result: str, ex: int):
-        """异步设置任务结果"""
-        await self.async_redis.hset(key, "result", result)
-        return await self.async_redis.expire(key, ex)
-    
     async def get_and_delayed_deletion(self, key: str, ex: int):
         """获取结果并延迟删除（从hash中）"""
         result = await self.async_redis.hget(key, "result")
         await self.async_redis.expire(key, ex)
         return result
-    
-    async def _get_result_async(self, key: str, delete: bool, delayed_deletion_ex: int):
-        """异步获取任务结果"""
-        client = self.async_redis
-        if delayed_deletion_ex is not None:
-            result = await client.hget(key, "result")
-            await client.expire(key, delayed_deletion_ex)
-            return result
-        elif delete:
-            # 获取结果并删除整个hash
-            result = await client.hget(key, "result")
-            await client.delete(key)
-            return result
-        else:
-            return await client.hget(key, "result")
-    
-    def get_result(self, event_id: str, delete: bool = False, asyncio: bool = False, 
-                   delayed_deletion_ex: int = None, wait: bool = False, timeout: int = 300,
-                   poll_interval: float = 0.5):
-        """获取任务结果（从TASK:hash的result字段）
-        
-        Args:
-            event_id: 任务ID
-            delete: 是否删除结果
-            asyncio: 是否使用异步模式
-            delayed_deletion_ex: 延迟删除时间（秒）
-            wait: 是否阻塞等待直到任务完成
-            timeout: 等待超时时间（秒），默认300秒
-            poll_interval: 轮询间隔（秒），默认0.5秒
-            suppress_traceback: 是否抑制框架层堆栈（直接打印错误并退出）
-        
-        Returns:
-            任务结果字符串
-            
-        Raises:
-            TaskTimeoutError: 等待超时
-            TaskExecutionError: 任务执行失败
-            TaskNotFoundError: 任务不存在
-        """
-        if asyncio:
-            key = f"{self.redis_prefix}:TASK:{event_id}"
-            if wait:
-                return self._get_result_async_wait(event_id, key, delete, delayed_deletion_ex, timeout, poll_interval)
-            else:
-                return self._get_result_async(key, delete, delayed_deletion_ex)
-        else:
-            # 同步模式
-            if wait:
-                return self._get_result_sync_wait(event_id, delete, delayed_deletion_ex, timeout, poll_interval)
-            else:
-                client = self.redis
-                key = f"{self.redis_prefix}:TASK:{event_id}"
-                if delayed_deletion_ex is not None:
-                    result = client.hget(key, "result")
-                    client.expire(key, delayed_deletion_ex)
-                    return result
-                elif delete:
-                    # 如果配置了任务中心，不删除消息，等任务中心同步后删除
-                    if self.task_center and self.task_center.is_enabled:
-                        result = client.hget(key, "result")
-                        # 仅标记为待删除，不实际删除
-                        client.hset(key, "__pending_delete", "1")
-                        return result
-                    else:
-                        # 获取结果并删除整个hash
-                        result = client.hget(key, "result")
-                        client.delete(key)
-                        return result
-                else:
-                    # 先尝试从Redis获取
-                    result = client.hget(key, "result")
-                    # 如果Redis中没有且配置了任务中心，从任务中心获取
-                    if result is None and self.task_center_client.is_enabled:
-                        import asyncio
-                        loop = asyncio.new_event_loop()
-                        try:
-                            task_data = loop.run_until_complete(
-                                self.task_center_client.get_task_result(event_id)
-                            )
-                            if task_data:
-                                result = task_data.get('result')
-                        finally:
-                            loop.close()
-                    return result
-    
-    def _get_result_sync_wait(self, event_id: str, delete: bool, delayed_deletion_ex: int, 
-                              timeout: int, poll_interval: float):
-        """同步模式下阻塞等待任务结果"""
-        start_time = time.time()
-        
-        while True:
-            # 检查超时
-            if time.time() - start_time > timeout:
-                raise TaskTimeoutError(f"Task {event_id} timed out after {timeout} seconds")
-            
-            # 获取任务状态
-            status = self.get_task_status(event_id)
-            
-            if status is None:
-                raise TaskNotFoundError(f"Task {event_id} not found")
-            
-            if status == 'success':
-                # 任务成功，获取结果
-                key = f"{self.redis_prefix}:TASK:{event_id}"
-                if delayed_deletion_ex is not None:
-                    result = self.redis.hget(key, "result")
-                    self.redis.expire(key, delayed_deletion_ex)
-                    return result
-                elif delete:
-                    result = self.redis.hget(key, "result")
-                    self.redis.delete(key)
-                    return result
-                else:
-                    return self.redis.hget(key, "result")
-            
-            elif status == 'error':
-                # 任务失败，获取错误信息并抛出异常
-                key = f"{self.redis_prefix}:TASK:{event_id}"
-                # 从 exception 字段获取错误信息
-                error_msg = self.redis.hget(key, "exception") or "Task execution failed"
-                # 抛出自定义异常
-                raise TaskExecutionError(event_id, error_msg)
-            
-            # 任务还在执行中，继续等待
-            time.sleep(poll_interval)
-    
-    async def _get_result_async_wait(self, event_id: str, key: str, delete: bool, 
-                                     delayed_deletion_ex: int, timeout: int, poll_interval: float):
-        """异步模式下等待任务结果"""
-        start_time = time.time()
-        
-        while True:
-            # 检查超时
-            if time.time() - start_time > timeout:
-                raise TaskTimeoutError(f"Task {event_id} timed out after {timeout} seconds")
-            
-            # 获取任务状态
-            status = await self._get_task_status_async(event_id)
-            
-            if status is None:
-                raise TaskNotFoundError(f"Task {event_id} not found")
-            
-            if status == 'success':
-                # 任务成功，获取结果
-                if delayed_deletion_ex is not None:
-                    result = await self.async_redis.hget(key, "result")
-                    await self.async_redis.expire(key, delayed_deletion_ex)
-                    return result
-                elif delete:
-                    result = await self.async_redis.hget(key, "result")
-                    await self.async_redis.delete(key)
-                    return result
-                else:
-                    return await self.async_redis.hget(key, "result")
-            
-            elif status == 'error':
-                # 任务失败，获取错误信息并抛出异常
-                # 从 exception 字段获取错误信息
-                error_msg = await self.async_redis.hget(key, "exception") or "Task execution failed"
-                # 抛出自定义异常
-                raise TaskExecutionError(event_id, error_msg)
-            
-            # 任务还在执行中，继续等待
-            await asyncio.sleep(poll_interval)
-    
+
     # ==================== 定时任务调度相关 ====================
     
     async def _ensure_scheduler_initialized(self, db_url: str = None):
         """确保调度器已初始化（内部方法）"""
         if not self.scheduler_manager:
-            logger.info("Auto-initializing scheduler...")
+            logger.debug("Auto-initializing scheduler...")
             # 优先使用传入的db_url，然后是实例化时的pg_url，最后是环境变量
             if not db_url:
                 db_url = self.pg_url or os.environ.get('JETTASK_PG_URL')
@@ -1356,7 +2262,8 @@ class Jettask(object):
                     "Or: export JETTASK_PG_URL='postgresql://user:password@localhost:5432/jettask'"
                 )
             
-            from ..scheduler import TaskScheduler, ScheduledTaskManager
+            from ..scheduler import TaskScheduler
+            from ..scheduler.task_crud import ScheduledTaskManager
             
             # 创建数据库管理器
             self.scheduler_manager = ScheduledTaskManager(db_url)
@@ -1376,7 +2283,7 @@ class Jettask(object):
             )
             
             await self.scheduler.connect()
-            logger.info("Scheduler initialized")
+            logger.debug("Scheduler initialized")
     
     async def start_scheduler(self):
         """启动定时任务调度器（自动初始化）"""
@@ -1393,7 +2300,7 @@ class Jettask(object):
         """停止定时任务调度器"""
         if self.scheduler:
             self.scheduler.stop()
-            logger.info("Scheduler stopped")
+            logger.debug("Scheduler stopped")
     
     async def add_scheduled_task(
         self,
@@ -1515,9 +2422,9 @@ class Jettask(object):
         task, created = await self.scheduler_manager.create_or_get_task(task, skip_if_exists=skip_if_exists)
         
         if created:
-            logger.info(f"Scheduled task {task.id} created for function {task_name}")
+            logger.debug(f"Scheduled task {task.id} created for function {task_name}")
         else:
-            logger.info(f"Scheduled task {task.id} already exists for function {task_name}")
+            logger.debug(f"Scheduled task {task.id} already exists for function {task_name}")
         
         return task
     
@@ -1627,7 +2534,7 @@ class Jettask(object):
         # 批量创建
         created_tasks = await self.scheduler_manager.batch_create_tasks(task_objects, skip_existing)
         
-        logger.info(f"Batch created {len(created_tasks)} scheduled tasks")
+        logger.debug(f"Batch created {len(created_tasks)} scheduled tasks")
         return created_tasks
     
     async def bulk_write_scheduled_tasks(self, tasks: list):
@@ -1676,7 +2583,7 @@ class Jettask(object):
         skip_existing = getattr(tasks[0], '_skip_if_exists', True) if tasks else True
         created_tasks = await self.scheduler_manager.batch_create_tasks(task_objects, skip_existing)
         
-        logger.info(f"Bulk wrote {len(created_tasks)} scheduled tasks")
+        logger.debug(f"Bulk wrote {len(created_tasks)} scheduled tasks")
         return created_tasks
     
     async def list_scheduled_tasks(self, **filters):
@@ -1709,7 +2616,7 @@ class Jettask(object):
             if self.scheduler:
                 await self.scheduler.loader.remove_task(task.id)
             
-            logger.info(f"Task {task.id} (scheduler_id: {task.scheduler_id}) disabled")
+            logger.debug(f"Task {task.id} (scheduler_id: {task.scheduler_id}) disabled")
             return True
         return False
     
@@ -1730,7 +2637,7 @@ class Jettask(object):
             if self.scheduler:
                 await self.scheduler.loader.load_tasks()
             
-            logger.info(f"Task {task.id} (scheduler_id: {task.scheduler_id}) enabled")
+            logger.debug(f"Task {task.id} (scheduler_id: {task.scheduler_id}) enabled")
             return True
         return False