@agentunion/kite 1.3.2 → 1.5.0

package/extensions/services/proxy/agentcp/monitoring/metrics_store.py

@@ -0,0 +1,325 @@
+# -*- coding: utf-8 -*-
+"""
+时间序列指标存储
+
+使用 SQLite 存储时间序列监控数据，支持历史查询和趋势分析
+"""
+
+import sqlite3
+import time
+import threading
+from typing import List, Dict, Any, Optional
+
+
+class MetricsStore:
+    """时间序列数据存储（基于SQLite）
+
+    特性：
+    - 轻量级，无需额外部署
+    - 支持高效的时间范围查询
+    - 自动清理过期数据
+    - 线程安全
+    """
+
+    def __init__(self, db_path: str):
+        """初始化时间序列存储
+
+        Args:
+            db_path: SQLite 数据库文件路径
+        """
+        self.db_path = db_path
+        self.lock = threading.Lock()
+        self._init_db()
+
+    def _init_db(self):
+        """初始化数据库表结构"""
+        with self.lock:
+            conn = sqlite3.connect(self.db_path)
+            cursor = conn.cursor()
+
+            # 创建时间序列表
+            cursor.execute('''
+                CREATE TABLE IF NOT EXISTS metrics_timeseries (
+                    timestamp INTEGER PRIMARY KEY,
+                    agent_id TEXT NOT NULL,
+                    received_total INTEGER DEFAULT 0,
+                    dispatched_success INTEGER DEFAULT 0,
+                    dispatched_failed INTEGER DEFAULT 0,
+                    handler_success INTEGER DEFAULT 0,
+                    handler_failed INTEGER DEFAULT 0,
+                    dispatch_queue_size INTEGER DEFAULT 0,
+                    avg_dispatch_latency_ms REAL DEFAULT 0.0,
+                    avg_handler_latency_ms REAL DEFAULT 0.0,
+                    p50_dispatch_latency_ms REAL DEFAULT 0.0,
+                    p95_dispatch_latency_ms REAL DEFAULT 0.0,
+                    p99_dispatch_latency_ms REAL DEFAULT 0.0,
+                    throughput_per_second REAL DEFAULT 0.0,
+                    success_rate REAL DEFAULT 0.0
+                )
+            ''')
+
+            # 创建索引（优化查询性能）
+            cursor.execute('''
+                CREATE INDEX IF NOT EXISTS idx_timestamp
+                ON metrics_timeseries(timestamp)
+            ''')
+            cursor.execute('''
+                CREATE INDEX IF NOT EXISTS idx_agent_id_timestamp
+                ON metrics_timeseries(agent_id, timestamp)
+            ''')
+
+            conn.commit()
+            conn.close()
+
+    def insert_snapshot(self, metrics: dict):
+        """插入一个时间点的指标快照（非阻塞）
+
+        Args:
+            metrics: 指标字典，必须包含以下字段：
+                - agent_id: AgentID 标识
+                - timestamp: 时间戳（可选，默认当前时间）
+                - received_total: 累计接收消息总数
+                - dispatched_success: 累计派发成功数
+                - dispatched_failed: 累计派发失败数
+                - handler_success: 累计处理成功数（可选）
+                - handler_failed: 累计处理失败数（可选）
+                - dispatch_queue_size: 当前派发队列大小
+                - avg_dispatch_latency_ms: 平均派发延迟（可选）
+                - avg_handler_latency_ms: 平均处理延迟（可选）
+                - p50_dispatch_latency_ms: P50延迟（可选）
+                - p95_dispatch_latency_ms: P95延迟（可选）
+                - p99_dispatch_latency_ms: P99延迟（可选）
+        """
+        timestamp = metrics.get('timestamp', int(time.time()))
+        if isinstance(timestamp, float):
+            timestamp = int(timestamp)
+
+        # ✅ 使用 trylock 模式：如果锁被占用，跳过本次写入（不阻塞）
+        locked = self.lock.acquire(blocking=False)
+        if not locked:
+            # 锁被占用，跳过本次写入（避免阻塞主流程）
+            return
+
+        try:
+            conn = sqlite3.connect(self.db_path, timeout=1.0)  # 1秒超时
+            cursor = conn.cursor()
+
+            try:
+                # 计算吞吐量和成功率
+                received_total = metrics.get('received_total', 0)
+                dispatched_success = metrics.get('dispatched_success', 0)
+                uptime = metrics.get('uptime_seconds', 1)
+
+                throughput = received_total / max(uptime, 1)
+                success_rate = (dispatched_success / max(received_total, 1)) * 100 if received_total > 0 else 0.0
+
+                cursor.execute('''
+                    INSERT OR REPLACE INTO metrics_timeseries (
+                        timestamp, agent_id, received_total, dispatched_success,
+                        dispatched_failed, handler_success, handler_failed,
+                        dispatch_queue_size, avg_dispatch_latency_ms, avg_handler_latency_ms,
+                        p50_dispatch_latency_ms, p95_dispatch_latency_ms, p99_dispatch_latency_ms,
+                        throughput_per_second, success_rate
+                    ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+                ''', (
+                    timestamp,
+                    metrics.get('agent_id', 'unknown'),
+                    received_total,
+                    dispatched_success,
+                    metrics.get('dispatched_failed', 0),
+                    metrics.get('handler_success', 0),
+                    metrics.get('handler_failed', 0),
+                    metrics.get('dispatch_queue_size', 0),
+                    self._safe_float(metrics.get('avg_dispatch_latency_ms')),
+                    self._safe_float(metrics.get('avg_handler_latency_ms')),
+                    self._safe_float(metrics.get('p50_dispatch_latency_ms')),
+                    self._safe_float(metrics.get('p95_dispatch_latency_ms')),
+                    self._safe_float(metrics.get('p99_dispatch_latency_ms')),
+                    throughput,
+                    success_rate,
+                ))
+
+                conn.commit()
+            except Exception as e:
+                print(f"❌ [MetricsStore] 插入数据失败: {e}")
+                conn.rollback()
+            finally:
+                conn.close()
+        finally:
+            # ✅ 确保释放锁（无论是否成功）
+            self.lock.release()
+
+    def _safe_float(self, value) -> float:
+        """安全地转换为浮点数
+
+        Args:
+            value: 待转换的值
+
+        Returns:
+            转换后的浮点数，失败返回0.0
+        """
+        if value is None:
+            return 0.0
+        if isinstance(value, str):
+            try:
+                return float(value)
+            except ValueError:
+                return 0.0
+        return float(value)
+
+    def query_range(
+        self,
+        from_ts: int,
+        to_ts: int,
+        agent_id: Optional[str] = None,
+        limit: int = 1000
+    ) -> List[Dict[str, Any]]:
+        """查询时间范围内的数据
+
+        Args:
+            from_ts: 起始时间戳
+            to_ts: 结束时间戳
+            agent_id: AgentID 过滤（可选）
+            limit: 最大返回记录数
+
+        Returns:
+            时间序列数据列表
+        """
+        with self.lock:
+            conn = sqlite3.connect(self.db_path)
+            cursor = conn.cursor()
+
+            try:
+                if agent_id:
+                    cursor.execute('''
+                        SELECT * FROM metrics_timeseries
+                        WHERE timestamp >= ? AND timestamp <= ? AND agent_id = ?
+                        ORDER BY timestamp ASC
+                        LIMIT ?
+                    ''', (from_ts, to_ts, agent_id, limit))
+                else:
+                    cursor.execute('''
+                        SELECT * FROM metrics_timeseries
+                        WHERE timestamp >= ? AND timestamp <= ?
+                        ORDER BY timestamp ASC
+                        LIMIT ?
+                    ''', (from_ts, to_ts, limit))
+
+                rows = cursor.fetchall()
+                return [self._row_to_dict(row) for row in rows]
+            finally:
+                conn.close()
+
+    def query_latest(self, agent_id: Optional[str] = None, limit: int = 100) -> List[Dict[str, Any]]:
+        """查询最新的数据点
+
+        Args:
+            agent_id: AgentID 过滤（可选）
+            limit: 最大返回记录数
+
+        Returns:
+            最新的时间序列数据列表
+        """
+        with self.lock:
+            conn = sqlite3.connect(self.db_path)
+            cursor = conn.cursor()
+
+            try:
+                if agent_id:
+                    cursor.execute('''
+                        SELECT * FROM metrics_timeseries
+                        WHERE agent_id = ?
+                        ORDER BY timestamp DESC
+                        LIMIT ?
+                    ''', (agent_id, limit))
+                else:
+                    cursor.execute('''
+                        SELECT * FROM metrics_timeseries
+                        ORDER BY timestamp DESC
+                        LIMIT ?
+                    ''', (limit,))
+
+                rows = cursor.fetchall()
+                return [self._row_to_dict(row) for row in rows]
+            finally:
+                conn.close()
+
+    def _row_to_dict(self, row) -> Dict[str, Any]:
+        """将数据库行转换为字典
+
+        Args:
+            row: 数据库查询结果行
+
+        Returns:
+            包含所有字段的字典
+        """
+        columns = [
+            'timestamp', 'agent_id', 'received_total', 'dispatched_success',
+            'dispatched_failed', 'handler_success', 'handler_failed',
+            'dispatch_queue_size', 'avg_dispatch_latency_ms', 'avg_handler_latency_ms',
+            'p50_dispatch_latency_ms', 'p95_dispatch_latency_ms', 'p99_dispatch_latency_ms',
+            'throughput_per_second', 'success_rate'
+        ]
+        return dict(zip(columns, row))
+
+    def cleanup_old_data(self, retention_days: int = 7):
+        """清理过期数据
+
+        Args:
+            retention_days: 数据保留天数（默认7天）
+        """
+        cutoff = int(time.time()) - (retention_days * 86400)
+
+        with self.lock:
+            conn = sqlite3.connect(self.db_path)
+            cursor = conn.cursor()
+
+            try:
+                cursor.execute('DELETE FROM metrics_timeseries WHERE timestamp < ?', (cutoff,))
+                deleted_count = cursor.rowcount
+                conn.commit()
+
+                if deleted_count > 0:
+                    print(f"🧹 [MetricsStore] 清理了 {deleted_count} 条过期数据 (>{retention_days}天)")
+            except Exception as e:
+                print(f"❌ [MetricsStore] 清理数据失败: {e}")
+                conn.rollback()
+            finally:
+                conn.close()
+
+    def get_stats(self) -> Dict[str, Any]:
+        """获取数据库统计信息
+
+        Returns:
+            包含数据库统计信息的字典
+        """
+        with self.lock:
+            conn = sqlite3.connect(self.db_path)
+            cursor = conn.cursor()
+
+            try:
+                # 查询总记录数
+                cursor.execute('SELECT COUNT(*) FROM metrics_timeseries')
+                total_records = cursor.fetchone()[0]
+
+                # 查询时间范围
+                cursor.execute('SELECT MIN(timestamp), MAX(timestamp) FROM metrics_timeseries')
+                min_ts, max_ts = cursor.fetchone()
+
+                # 查询不同 agent_id 数量
+                cursor.execute('SELECT COUNT(DISTINCT agent_id) FROM metrics_timeseries')
+                agent_count = cursor.fetchone()[0]
+
+                return {
+                    'total_records': total_records,
+                    'min_timestamp': min_ts,
+                    'max_timestamp': max_ts,
+                    'agent_count': agent_count,
+                    'db_path': self.db_path,
+                }
+            finally:
+                conn.close()
+
+    def close(self):
+        """关闭存储（预留接口，SQLite 会自动管理连接）"""
+        pass

package/extensions/services/proxy/agentcp/monitoring/monitoring_service.py

@@ -0,0 +1,269 @@
+# -*- coding: utf-8 -*-
+"""
+统一监控服务
+
+集成滑动窗口统计和时间序列存储，提供完整的监控解决方案
+"""
+
+import threading
+import time
+from typing import Dict, Any, List, Optional
+
+from .sliding_window import SlidingWindowMetrics
+from .metrics_store import MetricsStore
+
+
+class MonitoringService:
+    """统一监控服务
+
+    职责：
+    - 定期采集指标快照（每10秒）
+    - 更新滑动窗口统计
+    - 持久化到时间序列存储
+    - 提供实时和历史数据查询接口
+    """
+
+    def __init__(
+        self,
+        agent_id: str,
+        metrics_collector,
+        db_path: str,
+        snapshot_interval: int = 10
+    ):
+        """初始化监控服务
+
+        Args:
+            agent_id: AgentID 标识
+            metrics_collector: MessageMetrics 实例
+            db_path: 时间序列数据库路径
+            snapshot_interval: 快照间隔（秒），默认10秒
+        """
+        self.agent_id = agent_id
+        self.metrics_collector = metrics_collector
+        self.snapshot_interval = snapshot_interval
+
+        # 初始化组件
+        self.sliding_windows = SlidingWindowMetrics()
+        self.metrics_store = MetricsStore(db_path)
+
+        # 线程控制
+        self._running = False
+        self._snapshot_thread: Optional[threading.Thread] = None
+
+        # 统计信息
+        self._snapshot_count = 0
+        self._last_cleanup_time = time.time()
+
+    def start(self):
+        """启动监控服务"""
+        if self._running:
+            print(f"⚠️ [MonitoringService] 监控服务已在运行")
+            return
+
+        self._running = True
+        self._snapshot_thread = threading.Thread(
+            target=self._snapshot_loop,
+            daemon=True,
+            name=f"MetricsSnapshot-{self.agent_id}"
+        )
+        self._snapshot_thread.start()
+        print(f"📊 [MonitoringService] 已启动 (agent_id={self.agent_id}, interval={self.snapshot_interval}s)")
+
+    def stop(self, wait: bool = True):
+        """停止监控服务
+
+        Args:
+            wait: 是否等待线程完全停止（默认True）
+                  设为False可避免阻塞主流程
+        """
+        if not self._running:
+            return
+
+        self._running = False
+
+        # 如果需要等待线程停止
+        if wait and self._snapshot_thread and self._snapshot_thread.is_alive():
+            self._snapshot_thread.join(timeout=5.0)
+
+        # 最后一次快照保存（仅在等待模式下执行）
+        if wait:
+            try:
+                self._take_snapshot()
+                print(f"📊 [MonitoringService] 已停止 (共采集 {self._snapshot_count} 次快照)")
+            except Exception as e:
+                print(f"⚠️ [MonitoringService] 最终快照失败: {e}")
+        else:
+            print(f"📊 [MonitoringService] 停止信号已发送（非阻塞模式）")
+
+        if wait:
+            self._snapshot_thread = None
+
+    def _snapshot_loop(self):
+        """快照循环 - 每N秒收集一次数据"""
+        while self._running:
+            try:
+                self._take_snapshot()
+
+                # 定期清理旧数据（每小时一次）
+                now = time.time()
+                if now - self._last_cleanup_time > 3600:
+                    self._cleanup_old_data()
+                    self._last_cleanup_time = now
+
+            except Exception as e:
+                print(f"❌ [MonitoringService] 快照失败: {e}")
+                import traceback
+                traceback.print_exc()
+
+            # 等待下一次快照
+            time.sleep(self.snapshot_interval)
+
+    def _take_snapshot(self):
+        """执行一次快照采集（非阻塞）"""
+        try:
+            # 1. 获取当前指标（使用 timeout 防止阻塞）
+            current_metrics = self.metrics_collector.get_summary()
+            current_metrics['agent_id'] = self.agent_id
+            current_metrics['timestamp'] = time.time()
+        except Exception as e:
+            # 获取指标失败，跳过本次快照（不影响核心流程）
+            print(f"⚠️ [MonitoringService] 获取指标失败，跳过本次快照: {e}")
+            return
+
+        # 解析字符串格式的指标（兼容现有 MessageMetrics）
+        self._parse_metrics(current_metrics)
+
+        # 2. 更新滑动窗口
+        self.sliding_windows.update(current_metrics)
+
+        # 3. 存储到时间序列数据库
+        self.metrics_store.insert_snapshot(current_metrics)
+
+        self._snapshot_count += 1
+
+        # 调试日志（可选）
+        if self._snapshot_count % 6 == 0:  # 每1分钟打印一次
+            print(
+                f"📊 [MonitoringService] 快照 #{self._snapshot_count}: "
+                f"received={current_metrics.get('received_total', 0)}, "
+                f"queue={current_metrics.get('dispatch_queue_size', 0)}"
+            )
+
+    def _parse_metrics(self, metrics: dict):
+        """解析和标准化指标数据
+
+        处理现有 MessageMetrics 返回的字符串格式指标
+        """
+        # 解析延迟数据（从嵌套字典中提取）
+        if 'dispatch_latency' in metrics and isinstance(metrics['dispatch_latency'], dict):
+            dispatch_latency = metrics['dispatch_latency']
+            metrics['avg_dispatch_latency_ms'] = self._parse_float(dispatch_latency.get('avg_ms', '0'))
+            metrics['p50_dispatch_latency_ms'] = self._parse_float(dispatch_latency.get('p50_ms', '0'))
+            metrics['p95_dispatch_latency_ms'] = self._parse_float(dispatch_latency.get('p95_ms', '0'))
+            metrics['p99_dispatch_latency_ms'] = self._parse_float(dispatch_latency.get('p99_ms', '0'))
+
+        if 'handler_latency' in metrics and isinstance(metrics['handler_latency'], dict):
+            handler_latency = metrics['handler_latency']
+            metrics['avg_handler_latency_ms'] = self._parse_float(handler_latency.get('avg_ms', '0'))
+
+        # 解析运行时间
+        if 'uptime_seconds' in metrics:
+            metrics['uptime_seconds'] = self._parse_float(metrics['uptime_seconds'])
+
+    def _parse_float(self, value) -> float:
+        """安全地解析浮点数"""
+        if isinstance(value, (int, float)):
+            return float(value)
+        if isinstance(value, str):
+            try:
+                return float(value)
+            except ValueError:
+                return 0.0
+        return 0.0
+
+    def _cleanup_old_data(self):
+        """清理旧数据"""
+        try:
+            self.metrics_store.cleanup_old_data(retention_days=7)
+        except Exception as e:
+            print(f"⚠️ [MonitoringService] 清理旧数据失败: {e}")
+
+    def get_realtime_metrics(self) -> Dict[str, Any]:
+        """获取实时指标（包括所有时间窗口）
+
+        Returns:
+            包含以下内容的字典：
+            - agent_id: AgentID 标识
+            - timestamp: 当前时间戳
+            - cumulative: 累计指标（来自 MessageMetrics）
+            - windows: 所有时间窗口的统计数据
+        """
+        base_metrics = self.metrics_collector.get_summary()
+        window_stats = self.sliding_windows.get_all_windows()
+
+        return {
+            'agent_id': self.agent_id,
+            'timestamp': time.time(),
+            'cumulative': base_metrics,  # 累计指标
+            'windows': window_stats,     # 时间窗口指标
+        }
+
+    def get_window_metrics(self, window_names: List[str]) -> Dict[str, Dict[str, Any]]:
+        """获取指定时间窗口的指标
+
+        Args:
+            window_names: 窗口名称列表，如 ['1m', '3m', '5m']
+
+        Returns:
+            窗口统计数据字典
+        """
+        all_windows = self.sliding_windows.get_all_windows()
+        return {
+            name: all_windows.get(name, {})
+            for name in window_names
+            if name in all_windows
+        }
+
+    def get_history(self, from_ts: int, to_ts: int, limit: int = 1000) -> List[Dict[str, Any]]:
+        """获取历史数据
+
+        Args:
+            from_ts: 起始时间戳
+            to_ts: 结束时间戳
+            limit: 最大返回记录数
+
+        Returns:
+            历史数据列表
+        """
+        return self.metrics_store.query_range(from_ts, to_ts, self.agent_id, limit)
+
+    def get_latest_history(self, limit: int = 100) -> List[Dict[str, Any]]:
+        """获取最新的历史数据
+
+        Args:
+            limit: 最大返回记录数
+
+        Returns:
+            最新的历史数据列表
+        """
+        return self.metrics_store.query_latest(self.agent_id, limit)
+
+    def get_service_info(self) -> Dict[str, Any]:
+        """获取监控服务信息
+
+        Returns:
+            包含服务状态和统计信息的字典
+        """
+        store_stats = self.metrics_store.get_stats()
+
+        return {
+            'agent_id': self.agent_id,
+            'running': self._running,
+            'snapshot_interval': self.snapshot_interval,
+            'snapshot_count': self._snapshot_count,
+            'store_stats': store_stats,
+        }
+
+    def reset_windows(self):
+        """重置所有时间窗口（用于测试）"""
+        self.sliding_windows.reset()