fix(proxy): 抑制 tunnel/hub 高频重复日志，增加快速断开退避机制

- proxy_tunnel: idle timeout 日志首次 warning，后续降级为 debug/info 汇总
- hub_transport: 连续断开日志降级，追踪快速断开并调整重连初始 delay
- tunnel_manager: connect/disconnect 日志按 reconnect 计数分级输出

src/api/admin/proxy_tunnel.py

@@ -26,6 +26,9 @@ router = APIRouter()
 # Per-node 锁: 防止并发的 connect/disconnect 写入 DB 时出现竞态（后断连覆盖先连接）
 _node_status_locks: dict[str, asyncio.Lock] = {}
 
+# Per-node idle timeout 连续计数: 用于抑制重复日志
+_idle_timeout_counts: dict[str, int] = {}
+
 
 def _get_node_lock(node_id: str) -> asyncio.Lock:
     lock = _node_status_locks.get(node_id)
@@ -195,7 +198,19 @@ async def proxy_tunnel_ws(ws: WebSocket) -> None:
             try:
                 data = await asyncio.wait_for(ws.receive_bytes(), timeout=_IDLE_TIMEOUT)
             except asyncio.TimeoutError:
-                logger.warning("tunnel idle timeout for node_id={}", node_id)
+                count = _idle_timeout_counts.get(node_id, 0) + 1
+                _idle_timeout_counts[node_id] = count
+                # 首次 warning，后续每 10 次打印一条 info，其余 debug
+                if count == 1:
+                    logger.warning("tunnel idle timeout for node_id={}", node_id)
+                elif count % 10 == 0:
+                    logger.info(
+                        "tunnel idle timeout for node_id={} (repeated {} times)",
+                        node_id,
+                        count,
+                    )
+                else:
+                    logger.debug("tunnel idle timeout for node_id={}", node_id)
                 disconnect_reason = "idle timeout"
                 await ws.close(code=4004, reason="idle timeout")
                 break
@@ -209,6 +224,8 @@ async def proxy_tunnel_ws(ws: WebSocket) -> None:
                     break
                 continue
             oversized_count = 0  # 正常帧重置计数
+            _idle_timeout_counts.pop(node_id, None)  # 收到正常帧，重置 idle timeout 计数
+            manager.reset_reconnect_count(node_id)  # 连接正常活跃，重置 reconnect 计数
             try:
                 frame = Frame.decode(data)
             except ValueError as e:

src/services/proxy_node/hub_transport.py

@@ -9,6 +9,7 @@ from __future__ import annotations
 import asyncio
 import gzip
 import json
+import time as _time
 from datetime import datetime, timezone
 from typing import TYPE_CHECKING, Any
 
@@ -68,6 +69,11 @@ class HubConnectionManager:
 
         self._closing = False
 
+        self._disconnect_count = 0  # 连续断开计数，用于抑制重复日志
+        # 连续快速断开退避：防止 Hub 端持续发送 GOAWAY 时产生重连风暴
+        self._last_disconnect_ts: float = 0.0
+        self._rapid_disconnect_count: int = 0
+
     @property
     def is_connected(self) -> bool:
         ws = self._ws
@@ -126,7 +132,15 @@ class HubConnectionManager:
 
         self._reader_task = asyncio.create_task(self._reader_loop(ws))
         self._ping_task = asyncio.create_task(self._ping_loop(ws))
-        logger.info("Hub worker channel connected: {}", self._config.worker_ws_url)
+        if self._disconnect_count == 0:
+            logger.info("Hub worker channel connected: {}", self._config.worker_ws_url)
+        else:
+            logger.debug(
+                "Hub worker channel connected: {} (after {} disconnects)",
+                self._config.worker_ws_url,
+                self._disconnect_count,
+            )
+        self._disconnect_count = 0
 
     def _start_reconnect_loop(self) -> None:
         if self._closing:
@@ -138,7 +152,9 @@ class HubConnectionManager:
         self._reconnect_task = asyncio.create_task(self._reconnect_loop())
 
     async def _reconnect_loop(self) -> None:
-        attempt = 0
+        # 如果连续快速断开（GOAWAY 风暴），初始 attempt 跳过 0-delay 阶段
+        rapid = self._rapid_disconnect_count
+        attempt = min(rapid, len(_RECONNECT_DELAYS_SECONDS) - 1)
         while not self._closing and not self.is_connected:
             delay = _RECONNECT_DELAYS_SECONDS[min(attempt, len(_RECONNECT_DELAYS_SECONDS) - 1)]
             if delay > 0:
@@ -149,7 +165,7 @@ class HubConnectionManager:
                         break
                     await self._connect_once()
                 if self.is_connected:
-                    logger.info("Hub worker channel reconnected")
+                    logger.debug("Hub worker channel reconnected (attempt {})", attempt + 1)
                     break
             except Exception as e:
                 attempt += 1
@@ -192,7 +208,26 @@ class HubConnectionManager:
             self._pending_streams.clear()
 
         if not self._closing:
-            logger.warning("Hub worker channel disconnected: {}", reason)
+            self._disconnect_count += 1
+
+            # 追踪连续快速断开：如果距上次断开不足 2 秒，累加计数；否则重置
+            now_mono = _time.monotonic()
+            if now_mono - self._last_disconnect_ts < 2.0:
+                self._rapid_disconnect_count += 1
+            else:
+                self._rapid_disconnect_count = 0
+            self._last_disconnect_ts = now_mono
+
+            if self._disconnect_count <= 1:
+                logger.warning("Hub worker channel disconnected: {}", reason)
+            elif self._disconnect_count % 10 == 0:
+                logger.info(
+                    "Hub worker channel disconnected: {} (repeated {} times)",
+                    reason,
+                    self._disconnect_count,
+                )
+            else:
+                logger.debug("Hub worker channel disconnected: {}", reason)
             self._start_reconnect_loop()
 
     async def _send_frame(self, frame: Frame) -> None:

src/services/proxy_node/tunnel_manager.py

@@ -229,6 +229,8 @@ class TunnelManager:
         self._connections: dict[str, list[TunnelConnection]] = {}  # node_id -> [conn, ...]
         self._background_tasks: set[asyncio.Task[None]] = set()
         self._draining: bool = False
+        # 每个 node 的连续 reconnect 计数，用于抑制高频 connect/disconnect 日志
+        self._reconnect_counts: dict[str, int] = {}
 
     def _background(self, coro: Any) -> None:  # noqa: ANN401
         """启动 fire-and-forget task，通过 set 持有引用防止 GC 回收，完成后自动清理"""
@@ -269,12 +271,22 @@ class TunnelManager:
             conns = []
             self._connections[conn.node_id] = conns
         conns.append(conn)
-        logger.info(
-            "tunnel connected: node_id={}, name={}, pool_size={}",
-            conn.node_id,
-            conn.node_name,
-            len(conns),
-        )
+        reconn = self._reconnect_counts.get(conn.node_id, 0)
+        if reconn == 0:
+            logger.info(
+                "tunnel connected: node_id={}, name={}, pool_size={}",
+                conn.node_id,
+                conn.node_name,
+                len(conns),
+            )
+        else:
+            logger.debug(
+                "tunnel reconnected: node_id={}, name={}, pool_size={}, reconnect_count={}",
+                conn.node_id,
+                conn.node_name,
+                len(conns),
+                reconn,
+            )
 
     def unregister(self, conn: TunnelConnection) -> bool:
         """
@@ -297,14 +309,37 @@ class TunnelManager:
             self._connections.pop(conn.node_id, None)
 
         remaining = len(conns) if conns else 0
-        logger.info(
-            "tunnel disconnected: node_id={}, name={}, remaining={}",
-            conn.node_id,
-            conn.node_name,
-            remaining,
-        )
+        reconn = self._reconnect_counts.get(conn.node_id, 0) + 1
+        self._reconnect_counts[conn.node_id] = reconn
+        # 首次断开 info，后续每 10 次 info 汇总，其余 debug
+        if reconn == 1:
+            logger.info(
+                "tunnel disconnected: node_id={}, name={}, remaining={}",
+                conn.node_id,
+                conn.node_name,
+                remaining,
+            )
+        elif reconn % 10 == 0:
+            logger.info(
+                "tunnel disconnected: node_id={}, name={}, remaining={} (repeated {} times)",
+                conn.node_id,
+                conn.node_name,
+                remaining,
+                reconn,
+            )
+        else:
+            logger.debug(
+                "tunnel disconnected: node_id={}, name={}, remaining={}",
+                conn.node_id,
+                conn.node_name,
+                remaining,
+            )
         return True
 
+    def reset_reconnect_count(self, node_id: str) -> None:
+        """重置 reconnect 计数（当连接恢复正常活动时调用）"""
+        self._reconnect_counts.pop(node_id, None)
+
     async def shutdown_all(self, drain_timeout: float = 60.0) -> None:
         """优雅关闭所有 tunnel 连接：drain 飞行中请求 -> GoAway -> 关闭 WebSocket。