v2.2.0

约 1037 字大约 3 分钟

2026-05-19

日期：2026-05-18 类型：Feature / Architecture / Resilience

变更概述

针对工业通讯网关在恶劣网络环境（如切换 WIFI、短暂断网、网线松动等）下的稳定性要求，本次更新对 UniCon 驱动系统进行了连接自愈与网络韧性的深度升级：

全局网络状态监控 (INetworkMonitor)：引入硬件级网络状态变更感知，规避套接字断开事件的“虚假存活”迟滞。
底层事件驱动断线感知：各协议驱动全面挂载底层客户端事件（MQTT Disconnected, OPC UA Faulted/Closed 等），从“轮询心跳检测”升级为“主动事件触发”的毫秒级故障响应。
故障自愈 Watchdog 智能流控：结合网络可用性，在断网时智能挂起重连定时器，降低 CPU 开销，并在网络恢复瞬间实现毫秒级自动重连自愈。
全套接字异常防护降级：S7 和 Modbus 等传统无事件协议，在传输层捕获 any I/O 或套接字异常时即刻驱动状态机流转为 Faulted。

INetworkMonitor & NetworkMonitor: Added cross-platform system-level network status monitoring leveraging .NET's NetworkChange API.
DriverBase Self-Healing Watchdog: Upgraded base driver class to reactively pause reconnection schedules under a network outage and immediately resume/wakeup on network restoration.
MqttDriver: Integrated reactive connection monitoring by hooking the underlying client's DisconnectedAsync event.
OpcUaDriver: Integrated CommunicationState transition checks and hooked Session Faulted/Closed channels to instantly transit driver to Faulted state.
OpcUaPubSubDriver: Added ConnectionLost events to transport layers (MqttPubSubTransport and UdpPubSubTransport) to instantly react to protocol/socket exceptions.
S7Driver & ModbusDriver: Instrumented read, write, and batch communication scopes to transition the driver state to Faulted immediately upon physical client/socket loss.
Resilience Tests: Implemented a comprehensive suite of unit tests in ResilienceTests.cs validating network-down pausing and network-restored instant wakeup behaviors.

Reduced reconnect recovery latency under restored networking to milliseconds.
Reduced host machine CPU, socket port, and network thrashing during extended outages via passive pause mechanism.
Handled all event subscriptions cleanly in Dispose scopes to prevent memory leaks.

------details-----

实现全局网络状态感知并深度整合至所有驱动中，使驱动能够监听物理和协议客户端事件（Mqtt, OpcUa, Modbus, S7, OpcUaPubSub），保障在断网/网络切换时的秒级故障自愈，且在断网期间挂起重连以免资源浪费。

Step 1: 编写设计文档 connection-and-network-resilience.md 并获得架构师确认。
Step 2: 实现 INetworkMonitor 及其基于 NetworkChange 的系统级网络检测实现。
Step 3: 在 ServiceCollectionExtensions.cs 中向 IoC 注册 INetworkMonitor，并将其注入 DriverBase.cs。
Step 4: 重构 DriverBase.cs 的重连自愈 Watchdog 逻辑，当网络不可用时挂起重连线程，在网络恢复时即刻触发自愈重连。
Step 5: 为 IPubSubTransport 扩展 ConnectionLost 事件，并在 MqttPubSubTransport 和 UdpPubSubTransport 中实现。
Step 6: 逐一重构 5 大驱动，在连接建立时挂载各底层连接/通道/传输事件，并在事件触发或读写 I/O 抛出异常时令 State = DriverState.Faulted。
Step 7: 编写 ResilienceTests.cs 单元测试，并在本地运行 dotnet test 及 dotnet format 以确保质量门禁完全通过。
Step 8: 更新版本日志并结项归档。

------details end------