jettask 0.2.15__py3-none-any.whl → 0.2.17__py3-none-any.whl

jettask/services/overview_service.py

@@ -0,0 +1,978 @@
+"""
+概览服务层
+处理系统概览和健康检查相关的业务逻辑
+"""
+from datetime import datetime, timedelta, timezone
+from typing import Dict, Any, List, Optional
+import logging
+import time
+import traceback
+from sqlalchemy import text
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from jettask.schemas import TimeRangeQuery
+from jettask.core.db_manager import get_db_manager
+
+logger = logging.getLogger(__name__)
+
+# 获取全局数据访问实例
+# 使用新的统一数据库管理器
+
+
+class TimeRangeResult:
+    """时间范围处理结果"""
+    def __init__(self, start_time: datetime, end_time: datetime, interval: str, interval_seconds: int, granularity: str):
+        self.start_time = start_time
+        self.end_time = end_time
+        self.interval = interval
+        self.interval_seconds = interval_seconds
+        self.granularity = granularity
+
+
+class OverviewService:
+    """概览服务类"""
+    
+    @staticmethod
+    def get_root_info() -> Dict[str, Any]:
+        """
+        获取根路径信息
+        
+        Returns:
+            API基本信息
+        """
+        return {
+            "message": "JetTask Monitor API",
+            "version": "1.0.0",
+            "timestamp": datetime.now(timezone.utc).isoformat()
+        }
+    
+    @staticmethod
+    def get_health_status() -> Dict[str, Any]:
+        """
+        获取健康状态
+        
+        Returns:
+            健康状态信息
+        """
+        return {
+            "status": "healthy",
+            "timestamp": datetime.now(timezone.utc).isoformat()
+        }
+    
+    @staticmethod
+    async def get_system_stats(namespace: str) -> Dict[str, Any]:
+        """
+        获取指定命名空间的系统统计信息
+        
+        Args:
+            namespace: 命名空间名称
+            
+        Returns:
+            系统统计信息
+        """
+        db_manager = get_db_manager()
+        pool = await db_manager.get_pool(namespace)
+        redis_client = await pool.get_redis_client()
+        
+        try:
+            # 统计各种类型的键
+            stats = {
+                'namespace': namespace,
+                'queues': 0,
+                'tasks': 0,
+                'delayed_tasks': 0,
+                'workers': 0
+            }
+            
+            # 统计队列数量
+            queue_pattern = f"{namespace}:QUEUE:*"
+            async for _ in redis_client.scan_iter(match=queue_pattern):
+                stats['queues'] += 1
+            
+            # 统计任务数量
+            task_pattern = f"{namespace}:TASK:*"
+            async for _ in redis_client.scan_iter(match=task_pattern):
+                stats['tasks'] += 1
+            
+            # 统计延迟任务数量
+            delayed_pattern = f"{namespace}:DELAYED_QUEUE:*"
+            async for key in redis_client.scan_iter(match=delayed_pattern):
+                count = await redis_client.zcard(key)
+                stats['delayed_tasks'] += count
+            
+            # 统计工作进程数量
+            worker_pattern = f"{namespace}:WORKER:*"
+            async for _ in redis_client.scan_iter(match=worker_pattern):
+                stats['workers'] += 1
+            
+            return stats
+            
+        finally:
+            await redis_client.aclose()
+    
+    @staticmethod
+    async def get_dashboard_stats(
+        namespace: str,
+        time_range: str = "24h",
+        queues: Optional[str] = None
+    ) -> Dict[str, Any]:
+        """
+        获取仪表板统计数据
+        
+        Args:
+            namespace: 命名空间名称
+            time_range: 时间范围
+            queues: 逗号分隔的队列名称列表
+            
+        Returns:
+            仪表板统计数据
+        """
+        db_manager = get_db_manager()
+        pool = await db_manager.get_pool(namespace)
+        
+        # 如果没有PostgreSQL配置，返回空数据
+        # 检查是否配置了PostgreSQL
+        if not pool.config.has_postgres():
+            return {
+                "success": True,
+                "data": _get_empty_dashboard_stats()
+            }
+        
+        # 计算时间范围
+        end_time = datetime.now(timezone.utc)
+        start_time = _parse_time_range(time_range, end_time)
+        
+        # 构建队列筛选条件
+        queue_filter, queue_list, queue_params = _build_queue_filter_and_params(queues)
+        
+        async with pool.get_sa_session() as session:
+            # 获取统计数据
+            stats_data = await _get_task_statistics(
+                session, namespace, start_time, end_time,
+                queue_filter, queue_params
+            )
+            
+            # 计算吞吐量
+            throughput = await _calculate_throughput(
+                session, namespace, queue_filter, queue_params
+            )
+            
+            # 获取任务分布数据
+            # 合并所有查询参数
+            distribution_params = {
+                'namespace': namespace,
+                'start_time': start_time,
+                'end_time': end_time,
+                **queue_params
+            }
+            distribution_data = await _get_task_distribution(
+                session, namespace, start_time, end_time,
+                queue_filter, distribution_params
+            )
+            
+            return {
+                "success": True,
+                "data": {
+                    **stats_data,
+                    "throughput": throughput,
+                    "distribution": distribution_data
+                }
+            }
+    
+    @staticmethod
+    async def get_top_queues(
+        namespace: str,
+        metric: str = "backlog",
+        limit: int = 10,
+        time_range: str = "24h",
+        queues: Optional[str] = None
+    ) -> Dict[str, Any]:
+        """
+        获取队列排行榜
+        
+        Args:
+            namespace: 命名空间名称
+            metric: 指标类型 (backlog/error)
+            limit: 返回的队列数量限制
+            time_range: 时间范围
+            queues: 逗号分隔的队列名称列表
+            
+        Returns:
+            队列排行榜数据
+        """
+        if metric == "backlog":
+            return await _get_top_backlog_queues(namespace, limit, time_range, queues)
+        elif metric == "error":
+            return await _get_top_error_queues(namespace, limit, time_range, queues)
+        else:
+            raise ValueError(f"不支持的指标类型: {metric}")
+    
+    @staticmethod
+    async def get_dashboard_overview_stats(
+        namespace: str,
+        query: TimeRangeQuery
+    ) -> Dict[str, Any]:
+        """
+        获取概览页面的统一统计数据
+        
+        Args:
+            namespace: 命名空间名称
+            query: 时间范围查询参数
+            
+        Returns:
+            统一的时间序列数据
+        """
+        db_manager = get_db_manager()
+        pool = await db_manager.get_pool(namespace)
+        
+        # 如果没有PostgreSQL配置，返回空数据
+        # 检查是否配置了PostgreSQL
+        if not pool.config.has_postgres():
+            return _get_empty_overview_stats()
+        
+        # 解析时间范围
+        time_range_result = _parse_time_range_query(query)
+        
+        # 构建队列筛选条件
+        # 使用 TimeRangeQuery 的 queues 字段（支持多队列列表）
+        queue_list = query.queues if hasattr(query, 'queues') and query.queues else None
+        queue_filter, _, queue_params = _build_queue_filter_and_params(queue_list)
+        
+        async with pool.get_sa_session() as session:
+            # 执行统一查询
+            result = await _execute_overview_query(
+                session, namespace, time_range_result,
+                queue_filter, queue_params
+            )
+            
+            # 格式化数据
+            return _format_overview_data(result, time_range_result.granularity)
+
+
+# ============ 辅助函数 ============
+
+def _parse_time_range(time_range: str, end_time: datetime) -> datetime:
+    """解析时间范围字符串"""
+    if time_range.endswith('m'):
+        minutes = int(time_range[:-1])
+        return end_time - timedelta(minutes=minutes)
+    elif time_range.endswith('h'):
+        hours = int(time_range[:-1])
+        return end_time - timedelta(hours=hours)
+    elif time_range.endswith('d'):
+        days = int(time_range[:-1])
+        return end_time - timedelta(days=days)
+    else:
+        return end_time - timedelta(hours=24)  # 默认24小时
+
+
+def _parse_time_range_query(query: TimeRangeQuery) -> TimeRangeResult:
+    """解析TimeRangeQuery对象"""
+    # TimeRangeQuery 有 start_time, end_time 和 interval 字段
+    # 但没有 time_range 字段
+    
+    # 如果有结束时间，使用它；否则使用当前时间
+    if query.end_time:
+        end_time = datetime.fromisoformat(query.end_time.replace('Z', '+00:00')) if isinstance(query.end_time, str) else query.end_time
+    else:
+        end_time = datetime.now(timezone.utc)
+    
+    # 如果有开始时间，使用它；否则基于interval或默认24小时
+    if query.start_time:
+        start_time = datetime.fromisoformat(query.start_time.replace('Z', '+00:00')) if isinstance(query.start_time, str) else query.start_time
+    else:
+        # 使用interval字段来计算开始时间，默认为24小时
+        interval = query.interval or "24h"
+        start_time = _parse_time_range(interval, end_time)
+    
+    return _calculate_dynamic_interval(start_time, end_time)
+
+
+def _calculate_dynamic_interval(start_time: datetime, end_time: datetime, target_points: int = 200) -> TimeRangeResult:
+    """根据时间范围动态计算合适的时间间隔"""
+    duration = (end_time - start_time).total_seconds()
+    ideal_interval_seconds = duration / target_points
+    
+    # 选择合适的间隔
+    intervals = [
+        (1, '1 seconds', 'second'),
+        (5, '5 seconds', 'second'),
+        (10, '10 seconds', 'second'),
+        (30, '30 seconds', 'second'),
+        (60, '1 minute', 'minute'),
+        (300, '5 minutes', 'minute'),
+        (600, '10 minutes', 'minute'),
+        (1800, '30 minutes', 'minute'),
+        (3600, '1 hour', 'hour'),
+        (21600, '6 hours', 'hour'),
+        (43200, '12 hours', 'hour'),
+        (86400, '1 day', 'day')
+    ]
+    
+    for seconds, interval_str, granularity in intervals:
+        if ideal_interval_seconds <= seconds:
+            return TimeRangeResult(start_time, end_time, interval_str, seconds, granularity)
+    
+    # 默认返回1天间隔
+    return TimeRangeResult(start_time, end_time, '1 day', 86400, 'day')
+
+
+def _build_queue_filter_and_params(queues: Optional[List[str]] = None):
+    """构建队列筛选条件和参数"""
+    queue_list = []
+    if queues:
+        # 如果是字符串，按逗号分割（向后兼容）
+        if isinstance(queues, str):
+            queue_list = [q.strip() for q in queues.split(',') if q.strip()]
+        # 如果是列表，直接使用
+        elif isinstance(queues, list):
+            queue_list = [q.strip() for q in queues if q and q.strip()]
+        else:
+            queue_list = []
+    
+    queue_filter = ""
+    queue_params = {}
+    
+    if queue_list:
+        queue_placeholders = ','.join([f':queue_{i}' for i in range(len(queue_list))])
+        queue_filter = f"AND t.queue IN ({queue_placeholders})"
+        
+        for i, queue in enumerate(queue_list):
+            queue_params[f'queue_{i}'] = queue
+    
+    return queue_filter, queue_list, queue_params
+
+
+def _get_empty_dashboard_stats():
+    """返回空的仪表板统计数据"""
+    return {
+        "total_tasks": 0,
+        "completed_tasks": 0,
+        "failed_tasks": 0,
+        "running_tasks": 0,
+        "pending_tasks": 0,
+        "success_rate": 0,
+        "throughput": 0,
+        "avg_processing_time": 0,
+        "total_queues": 0,
+        "distribution": [{'type': '暂无数据', 'value': 1, 'queue': '', 'status': 'empty'}]
+    }
+
+
+def _get_empty_overview_stats():
+    """返回空的概览统计数据"""
+    return {
+        "task_trend": [],
+        "concurrency": [],
+        "processing_time": [],
+        "creation_latency": [],
+        "granularity": "minute"
+    }
+
+
+async def _get_task_statistics(session, namespace, start_time, end_time, queue_filter, queue_params):
+    """获取任务统计数据"""
+    stats_sql = text(f"""
+        WITH task_stats AS (
+            SELECT 
+                t.stream_id,
+                t.created_at,
+                t.queue,
+                tr.status,
+                tr.execution_time,
+                tr.end_time
+            FROM tasks t
+            LEFT JOIN task_runs tr ON t.stream_id = tr.stream_id
+            WHERE t.namespace = :namespace
+                AND t.created_at >= :start_time
+                AND t.created_at <= :end_time
+                {queue_filter}
+        )
+        SELECT 
+            COUNT(DISTINCT stream_id) as total_tasks,
+            COUNT(DISTINCT CASE WHEN status = 'success' THEN stream_id END) as completed_tasks,
+            COUNT(DISTINCT CASE WHEN status = 'error' THEN stream_id END) as failed_tasks,
+            COUNT(DISTINCT CASE WHEN status = 'running' THEN stream_id END) as running_tasks,
+            COUNT(DISTINCT CASE WHEN status IS NULL OR status = 'pending' THEN stream_id END) as pending_tasks,
+            COUNT(DISTINCT queue) as total_queues,
+            AVG(CASE WHEN status = 'success' AND execution_time IS NOT NULL 
+                THEN execution_time END) as avg_execution_time
+        FROM task_stats
+    """)
+    
+    # 合并所有查询参数
+    all_params = {
+        'namespace': namespace,
+        'start_time': start_time,
+        'end_time': end_time,
+        **queue_params
+    }
+    
+    result = await session.execute(stats_sql, all_params)
+    row = result.first()
+    
+    if row:
+        avg_execution_time = row.avg_execution_time or 0
+        success_rate = round((row.completed_tasks / row.total_tasks * 100) if row.total_tasks > 0 else 0, 1)
+        
+        return {
+            "total_tasks": row.total_tasks or 0,
+            "completed_tasks": row.completed_tasks or 0,
+            "failed_tasks": row.failed_tasks or 0,
+            "running_tasks": row.running_tasks or 0,
+            "pending_tasks": row.pending_tasks or 0,
+            "success_rate": success_rate,
+            "avg_processing_time": round(avg_execution_time * 1000 if avg_execution_time else 0, 1),
+            "total_queues": row.total_queues or 0
+        }
+    
+    return _get_empty_dashboard_stats()
+
+
+async def _calculate_throughput(session, namespace, queue_filter, queue_params):
+    """计算吞吐量（每分钟完成的任务数）"""
+    recent_end_time = datetime.now(timezone.utc)
+    throughput = 0
+    
+    time_windows = [
+        (5, "最近5分钟"),
+        (10, "最近10分钟"), 
+        (30, "最近30分钟"),
+        (60, "最近1小时")
+    ]
+    
+    for window_minutes, window_desc in time_windows:
+        recent_start_time = recent_end_time - timedelta(minutes=window_minutes)
+        
+        recent_query = text(f"""
+            SELECT COUNT(DISTINCT t.stream_id) as recent_completed
+            FROM tasks t
+            LEFT JOIN task_runs tr ON t.stream_id = tr.stream_id
+            WHERE t.namespace = :namespace
+                AND tr.status = 'success'
+                AND tr.end_time >= :recent_start_time
+                AND tr.end_time <= :recent_end_time
+                {queue_filter}
+        """)
+        
+        throughput_params = {
+            'namespace': namespace,
+            'recent_start_time': recent_start_time,
+            'recent_end_time': recent_end_time,
+            **queue_params
+        }
+        
+        recent_result = await session.execute(recent_query, throughput_params)
+        recent_row = recent_result.first()
+        recent_completed = recent_row.recent_completed if recent_row else 0
+        
+        if recent_completed >= 5:
+            throughput = round(recent_completed / window_minutes, 1)
+            logger.info(f"使用{window_desc}计算吞吐量: {recent_completed}个任务/{window_minutes}分钟 = {throughput}任务/分钟")
+            break
+        elif recent_completed > 0:
+            throughput = round(recent_completed / window_minutes, 1)
+    
+    return throughput
+
+
+async def _get_task_distribution(session, namespace, start_time, end_time, queue_filter, query_params):
+    """获取任务分布数据"""
+    distribution_sql = text(f"""
+        SELECT 
+            t.queue,
+            COUNT(DISTINCT t.stream_id) as count
+        FROM tasks t
+        WHERE t.namespace = :namespace
+            AND t.created_at >= :start_time
+            AND t.created_at <= :end_time
+            {queue_filter}
+        GROUP BY t.queue
+        ORDER BY count DESC, t.queue
+    """)
+    
+    distribution_result = await session.execute(distribution_sql, query_params)
+    
+    distribution_data = []
+    for row in distribution_result.fetchall():
+        if row.count > 0:
+            distribution_data.append({
+                'type': row.queue,
+                'value': row.count,
+                'queue': row.queue,
+                'status': 'all'
+            })
+    
+    if not distribution_data:
+        distribution_data = [
+            {'type': '暂无数据', 'value': 1, 'queue': '', 'status': 'empty'}
+        ]
+    
+    return distribution_data
+
+
+async def _get_top_backlog_queues(namespace, limit, time_range, queues):
+    """获取积压最多的队列Top N"""
+    db_manager = get_db_manager()
+    pool = await db_manager.get_pool(namespace)
+    
+    # 检查是否配置了PostgreSQL
+    if not pool.config.has_postgres():
+        return {"success": True, "data": []}
+    
+    end_time = datetime.now(timezone.utc)
+    start_time = _parse_time_range(time_range, end_time)
+    
+    async with pool.get_sa_session() as session:
+        queue_list = []
+        if queues:
+            # 如果是字符串，按逗号分割（向后兼容）
+            if isinstance(queues, str):
+                queue_list = [q.strip() for q in queues.split(',') if q.strip()]
+            # 如果是列表，直接使用
+            elif isinstance(queues, list):
+                queue_list = [q.strip() for q in queues if q and q.strip()]
+        
+        try:
+            # 优先从stream_backlog_monitor获取最新的积压数据
+            if queue_list:
+                backlog_sql = text("""
+                    SELECT 
+                        stream_name as queue,
+                        MAX(backlog_unprocessed) as backlog,
+                        CASE 
+                            WHEN MAX(backlog_unprocessed) > 100 THEN 'critical'
+                            WHEN MAX(backlog_unprocessed) > 50 THEN 'warning'
+                            ELSE 'normal'
+                        END as status
+                    FROM stream_backlog_monitor
+                    WHERE namespace = :namespace
+                        AND created_at >= :start_time
+                        AND created_at <= :end_time
+                        AND stream_name = ANY(:queues)
+                    GROUP BY stream_name
+                    HAVING MAX(backlog_unprocessed) > 0
+                    ORDER BY backlog DESC
+                    LIMIT :limit
+                """)
+                
+                result = await session.execute(backlog_sql, {
+                    'namespace': namespace,
+                    'start_time': start_time,
+                    'end_time': end_time,
+                    'queues': queue_list,
+                    'limit': limit
+                })
+            else:
+                backlog_sql = text("""
+                    SELECT 
+                        stream_name as queue,
+                        MAX(backlog_unprocessed) as backlog,
+                        CASE 
+                            WHEN MAX(backlog_unprocessed) > 100 THEN 'critical'
+                            WHEN MAX(backlog_unprocessed) > 50 THEN 'warning'
+                            ELSE 'normal'
+                        END as status
+                    FROM stream_backlog_monitor
+                    WHERE namespace = :namespace
+                        AND created_at >= :start_time
+                        AND created_at <= :end_time
+                    GROUP BY stream_name
+                    HAVING MAX(backlog_unprocessed) > 0
+                    ORDER BY backlog DESC
+                    LIMIT :limit
+                """)
+                
+                result = await session.execute(backlog_sql, {
+                    'namespace': namespace,
+                    'start_time': start_time,
+                    'end_time': end_time,
+                    'limit': limit
+                })
+            
+            backlog_queues = []
+            for row in result.fetchall():
+                backlog_queues.append({
+                    "queue": row.queue,
+                    "backlog": int(row.backlog),
+                    "status": row.status
+                })
+            
+            return {"success": True, "data": backlog_queues}
+            
+        except Exception as e:
+            logger.warning(f"从stream_backlog_monitor获取积压数据失败: {e}")
+            
+        # 如果失败，从tasks表统计
+        return await _get_top_backlog_from_tasks(session, namespace, limit)
+
+
+async def _get_top_backlog_from_tasks(session, namespace, limit):
+    """从tasks表获取积压数据"""
+    task_sql = text("""
+        SELECT 
+            t.queue,
+            COUNT(DISTINCT t.stream_id) as backlog,
+            CASE 
+                WHEN COUNT(DISTINCT t.stream_id) > 1000 THEN 'critical'
+                WHEN COUNT(DISTINCT t.stream_id) > 500 THEN 'warning'
+                ELSE 'normal'
+            END as status
+        FROM tasks t
+        LEFT JOIN task_runs tr ON t.stream_id = tr.stream_id
+        WHERE t.namespace = :namespace
+            AND (tr.stream_id IS NULL OR tr.status = 'pending')
+            AND t.created_at > NOW() - INTERVAL '24 hour'
+        GROUP BY t.queue
+        ORDER BY backlog DESC
+        LIMIT :limit
+    """)
+    
+    result = await session.execute(task_sql, {
+        'namespace': namespace,
+        'limit': limit
+    })
+    
+    backlog_queues = []
+    for row in result.fetchall():
+        backlog_queues.append({
+            "queue": row.queue,
+            "backlog": int(row.backlog),
+            "status": row.status
+        })
+    
+    return {"success": True, "data": backlog_queues}
+
+
+async def _get_top_error_queues(namespace, limit, time_range, queues):
+    """获取错误率最高的队列Top N"""
+    db_manager = get_db_manager()
+    pool = await db_manager.get_pool(namespace)
+    
+    # 检查是否配置了PostgreSQL
+    if not pool.config.has_postgres():
+        return {"success": True, "data": []}
+    
+    end_time = datetime.now(timezone.utc)
+    start_time = _parse_time_range(time_range, end_time)
+    
+    async with pool.get_sa_session() as session:
+        queue_list = []
+        if queues:
+            queue_list = [q.strip() for q in queues.split(',') if q.strip()]
+        
+        if queue_list:
+            error_sql = text("""
+                WITH queue_stats AS (
+                    SELECT 
+                        t.queue,
+                        COUNT(DISTINCT t.stream_id) as total,
+                        COUNT(DISTINCT CASE WHEN tr.status = 'error' THEN t.stream_id END) as errors
+                    FROM tasks t
+                    LEFT JOIN task_runs tr ON t.stream_id = tr.stream_id
+                    WHERE t.namespace = :namespace
+                        AND t.created_at >= :start_time
+                        AND t.created_at <= :end_time
+                        AND t.queue = ANY(:queues)
+                    GROUP BY t.queue
+                )
+                SELECT 
+                    queue,
+                    errors,
+                    total,
+                    CASE 
+                        WHEN total > 0 THEN ROUND(errors::numeric / total * 100, 1)
+                        ELSE 0 
+                    END as error_rate
+                FROM queue_stats
+                WHERE errors > 0
+                ORDER BY error_rate DESC, errors DESC
+                LIMIT :limit
+            """)
+            
+            result = await session.execute(error_sql, {
+                'namespace': namespace,
+                'start_time': start_time,
+                'end_time': end_time,
+                'queues': queue_list,
+                'limit': limit
+            })
+        else:
+            error_sql = text("""
+                WITH queue_stats AS (
+                    SELECT 
+                        t.queue,
+                        COUNT(DISTINCT t.stream_id) as total,
+                        COUNT(DISTINCT CASE WHEN tr.status = 'error' THEN t.stream_id END) as errors
+                    FROM tasks t
+                    LEFT JOIN task_runs tr ON t.stream_id = tr.stream_id
+                    WHERE t.namespace = :namespace
+                        AND t.created_at >= :start_time
+                        AND t.created_at <= :end_time
+                    GROUP BY t.queue
+                )
+                SELECT 
+                    queue,
+                    errors,
+                    total,
+                    CASE 
+                        WHEN total > 0 THEN ROUND(errors::numeric / total * 100, 1)
+                        ELSE 0 
+                    END as error_rate
+                FROM queue_stats
+                WHERE errors > 0
+                ORDER BY error_rate DESC, errors DESC
+                LIMIT :limit
+            """)
+            
+            result = await session.execute(error_sql, {
+                'namespace': namespace,
+                'start_time': start_time,
+                'end_time': end_time,
+                'limit': limit
+            })
+        
+        error_queues = []
+        for row in result.fetchall():
+            error_queues.append({
+                "queue": row.queue,
+                "errors": int(row.errors),
+                "total": int(row.total),
+                "error_rate": float(row.error_rate)
+            })
+        
+        return {"success": True, "data": error_queues}
+
+
+async def _execute_overview_query(session, namespace, time_range_result, queue_filter, queue_params):
+    """执行概览统计查询"""
+    interval_seconds = time_range_result.interval_seconds
+    
+    # 直接在SQL中使用间隔秒数，避免参数类型问题
+    sql = text(f"""
+        WITH time_series AS (
+            SELECT to_timestamp(FLOOR(EXTRACT(epoch FROM ts) / {interval_seconds}) * {interval_seconds}) AS time_bucket
+            FROM generate_series(
+                :start_time ::timestamptz,
+                :end_time ::timestamptz + INTERVAL '{interval_seconds} seconds',
+                INTERVAL '{interval_seconds} seconds'
+            ) AS ts
+        ),
+        enqueue_counts AS (
+            SELECT 
+                to_timestamp(FLOOR(EXTRACT(epoch FROM t.created_at) / {interval_seconds}) * {interval_seconds}) AS time_bucket,
+                COUNT(DISTINCT t.stream_id) as enqueued
+            FROM tasks t
+            WHERE t.namespace = :namespace
+                AND t.created_at >= :start_time
+                AND t.created_at <= :end_time
+                {queue_filter}
+            GROUP BY time_bucket
+        ),
+        complete_counts AS (
+            SELECT 
+                to_timestamp(FLOOR(EXTRACT(epoch FROM tr.end_time) / {interval_seconds}) * {interval_seconds}) AS time_bucket,
+                COUNT(DISTINCT t.stream_id) as completed
+            FROM task_runs tr
+            JOIN tasks t ON tr.stream_id = t.stream_id
+            WHERE t.namespace = :namespace
+                AND tr.end_time >= :start_time
+                AND tr.end_time <= :end_time
+                AND tr.status = 'success'
+                AND t.created_at >= :start_time
+                AND t.created_at <= :end_time
+                {queue_filter}
+            GROUP BY time_bucket
+        ),
+        failed_counts AS (
+            SELECT 
+                to_timestamp(FLOOR(EXTRACT(epoch FROM tr.end_time) / {interval_seconds}) * {interval_seconds}) AS time_bucket,
+                COUNT(DISTINCT t.stream_id) as failed
+            FROM task_runs tr
+            JOIN tasks t ON tr.stream_id = t.stream_id
+            WHERE t.namespace = :namespace
+                AND tr.end_time >= :start_time
+                AND tr.end_time <= :end_time
+                AND tr.status = 'error'
+                AND t.created_at >= :start_time
+                AND t.created_at <= :end_time
+                {queue_filter}
+            GROUP BY time_bucket
+        ),
+        concurrency_data AS (
+            SELECT 
+                to_timestamp(FLOOR(EXTRACT(epoch FROM tr.start_time) / {interval_seconds}) * {interval_seconds}) AS time_bucket,
+                COUNT(DISTINCT t.stream_id) as concurrent_tasks
+            FROM task_runs tr
+            JOIN tasks t ON tr.stream_id = t.stream_id
+            WHERE t.namespace = :namespace
+                AND tr.start_time >= :start_time
+                AND tr.start_time <= :end_time
+                AND tr.start_time IS NOT NULL
+                AND tr.end_time IS NOT NULL
+                AND t.created_at >= :start_time
+                AND t.created_at <= :end_time
+                {queue_filter}
+            GROUP BY time_bucket
+        ),
+        processing_time_data AS (
+            SELECT 
+                to_timestamp(FLOOR(EXTRACT(epoch FROM tr.end_time) / {interval_seconds}) * {interval_seconds}) AS time_bucket,
+                AVG(CASE WHEN tr.status = 'success' AND tr.execution_time > 0 
+                    THEN tr.execution_time END) as avg_processing_time,
+                PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY 
+                    CASE WHEN tr.status = 'success' AND tr.execution_time > 0 
+                    THEN tr.execution_time END) as p50_processing_time,
+                PERCENTILE_CONT(0.9) WITHIN GROUP (ORDER BY 
+                    CASE WHEN tr.status = 'success' AND tr.execution_time > 0 
+                    THEN tr.execution_time END) as p90_processing_time
+            FROM task_runs tr
+            JOIN tasks t ON tr.stream_id = t.stream_id
+            WHERE t.namespace = :namespace
+                AND tr.end_time >= :start_time
+                AND tr.end_time <= :end_time
+                AND tr.status = 'success'
+                {queue_filter}
+            GROUP BY time_bucket
+        ),
+        creation_latency_data AS (
+            SELECT 
+                to_timestamp(FLOOR(EXTRACT(epoch FROM tr.start_time) / {interval_seconds}) * {interval_seconds}) AS time_bucket,
+                AVG(EXTRACT(EPOCH FROM (tr.start_time - t.created_at))) as avg_creation_latency,
+                PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY 
+                    EXTRACT(EPOCH FROM (tr.start_time - t.created_at))) as p50_creation_latency,
+                PERCENTILE_CONT(0.9) WITHIN GROUP (ORDER BY 
+                    EXTRACT(EPOCH FROM (tr.start_time - t.created_at))) as p90_creation_latency
+            FROM task_runs tr
+            JOIN tasks t ON tr.stream_id = t.stream_id
+            WHERE t.namespace = :namespace
+                AND tr.start_time >= :start_time
+                AND tr.start_time <= :end_time
+                AND tr.start_time IS NOT NULL
+                {queue_filter}
+            GROUP BY time_bucket
+        )
+        SELECT 
+            ts.time_bucket,
+            COALESCE(eq.enqueued, 0) as enqueued,
+            COALESCE(cc.completed, 0) as completed,
+            COALESCE(fc.failed, 0) as failed,
+            COALESCE(cd.concurrent_tasks, 0) as concurrent_tasks,
+            ROUND(ptd.avg_processing_time::numeric, 6) as avg_processing_time,
+            ROUND(ptd.p50_processing_time::numeric, 6) as p50_processing_time,
+            ROUND(ptd.p90_processing_time::numeric, 6) as p90_processing_time,
+            ROUND(cld.avg_creation_latency::numeric, 3) as avg_creation_latency,
+            ROUND(cld.p50_creation_latency::numeric, 3) as p50_creation_latency,
+            ROUND(cld.p90_creation_latency::numeric, 3) as p90_creation_latency
+        FROM time_series ts
+        LEFT JOIN enqueue_counts eq ON ts.time_bucket = eq.time_bucket
+        LEFT JOIN complete_counts cc ON ts.time_bucket = cc.time_bucket
+        LEFT JOIN failed_counts fc ON ts.time_bucket = fc.time_bucket
+        LEFT JOIN concurrency_data cd ON ts.time_bucket = cd.time_bucket
+        LEFT JOIN processing_time_data ptd ON ts.time_bucket = ptd.time_bucket
+        LEFT JOIN creation_latency_data cld ON ts.time_bucket = cld.time_bucket
+        ORDER BY ts.time_bucket
+    """)
+    
+    query_params = {
+        'namespace': namespace,
+        'start_time': time_range_result.start_time,
+        'end_time': time_range_result.end_time,
+        **queue_params
+    }
+    
+    result = await session.execute(sql, query_params)
+    return result.fetchall()
+
+
+def _format_overview_data(rows, granularity):
+    """格式化概览数据"""
+    task_trend = []
+    concurrency = []
+    processing_time = []
+    creation_latency = []
+    
+    end_index = len(rows) - 1
+    
+    for idx, row in enumerate(rows):
+        time_str = row.time_bucket.isoformat()
+        
+        # 任务处理趋势数据
+        enqueued_val = row.enqueued if row.enqueued > 0 or idx == 0 or idx == end_index else None
+        completed_val = row.completed if row.completed > 0 or idx == 0 or idx == end_index else None
+        failed_val = row.failed if row.failed > 0 or idx == 0 or idx == end_index else None
+        
+        task_trend.extend([
+            {'time': time_str, 'value': enqueued_val, 'metric': '入队速率'},
+            {'time': time_str, 'value': completed_val, 'metric': '完成速率'},
+            {'time': time_str, 'value': failed_val, 'metric': '失败数'}
+        ])
+        
+        # 任务并发数量
+        concurrency.append({
+            'time': time_str,
+            'value': row.concurrent_tasks or 0,
+            'metric': '并发任务数'
+        })
+        
+        # 任务处理时间（转换为毫秒）
+        _add_processing_time_data(processing_time, time_str, row, idx, end_index)
+        
+        # 任务执行延时（秒）
+        _add_creation_latency_data(creation_latency, time_str, row, idx, end_index)
+    
+    return {
+        "task_trend": task_trend,
+        "concurrency": concurrency,
+        "processing_time": processing_time,
+        "creation_latency": creation_latency,
+        "granularity": granularity
+    }
+
+
+def _add_processing_time_data(processing_time, time_str, row, idx, end_index):
+    """添加处理时间数据"""
+    if row.avg_processing_time is not None:
+        avg_time_val = round(float(row.avg_processing_time * 1000), 1)
+    else:
+        avg_time_val = None if idx != 0 and idx != end_index else 0
+        
+    if row.p50_processing_time is not None:
+        p50_time_val = round(float(row.p50_processing_time * 1000), 1)
+    else:
+        p50_time_val = None if idx != 0 and idx != end_index else 0
+        
+    if row.p90_processing_time is not None:
+        p90_time_val = round(float(row.p90_processing_time * 1000), 1)
+    else:
+        p90_time_val = None if idx != 0 and idx != end_index else 0
+    
+    processing_time.extend([
+        {'time': time_str, 'value': avg_time_val, 'metric': '平均处理时间'},
+        {'time': time_str, 'value': p50_time_val, 'metric': 'P50处理时间'},
+        {'time': time_str, 'value': p90_time_val, 'metric': 'P90处理时间'}
+    ])
+
+
+def _add_creation_latency_data(creation_latency, time_str, row, idx, end_index):
+    """添加创建延迟数据"""
+    if row.avg_creation_latency is not None:
+        avg_latency_val = round(float(row.avg_creation_latency), 3)
+    else:
+        avg_latency_val = None if idx != 0 and idx != end_index else 0
+        
+    if row.p50_creation_latency is not None:
+        p50_latency_val = round(float(row.p50_creation_latency), 3)
+    else:
+        p50_latency_val = None if idx != 0 and idx != end_index else 0
+        
+    if row.p90_creation_latency is not None:
+        p90_latency_val = round(float(row.p90_creation_latency), 3)
+    else:
+        p90_latency_val = None if idx != 0 and idx != end_index else 0
+    
+    creation_latency.extend([
+        {'time': time_str, 'value': avg_latency_val, 'metric': '平均执行延时'},
+        {'time': time_str, 'value': p50_latency_val, 'metric': 'P50执行延时'},
+        {'time': time_str, 'value': p90_latency_val, 'metric': 'P90执行延时'}
+    ])