ASP.NET Core 6中PubNub服务运行数天后停止监听问题求助

问题描述 #

托管在IIS中的ASP.NET Core 6应用，使用PubNub 6.13.0订阅频道并处理消息。服务运行数天后会停止监听PubNub消息，但REST、SignalR等其他功能正常，仅能通过回收应用池恢复PubNub功能。该问题在20台服务器（AWS及本地）上均出现。

当前实现代码 #

服务注册 #

services.AddSingleton<IPubNubService, PubNubService>();

启动初始化 #

public void Configure(IApplicationBuilder app, IWebHostEnvironment env)
{
   //...
   app.ApplicationServices.GetService<IPubNubService>(); 
}

PubNubService简化实现 #

public class PubNubService : IPubNubService
{
    public PubNubService()
    {
        _pubNub = new Pubnub(new PNConfiguration(new UserId($"PubNub:UserId"))
        {
            SubscribeKey = config.GetValue<string>("PubNub:SubscribeKey"), 
            PublishKey = config.GetValue<string>("PubNub:PublishKey"), 
        });
        
        _pubNub.Subscribe<string>()
            .Channels(new[] { "PubNub:BroadcastChannel" })
            .Execute();

        _pubNub.AddListener(CreateListener());
    }

    private SubscribeCallbackExt CreateListener()
    {
        return new SubscribeCallbackExt
        (
            messageCallback: async delegate(Pubnub pnObj, PNMessageResult<object> pubMsg)
            {
                await HandleReceivedMessage(pubMsg);
            },
        );
    }

    private async Task HandleReceivedMessage(PNMessageResult<object> pubNubMessage)
    {
        // ... 消息处理逻辑
    }
}

故障时日志 #

DateTime 05/30/2023 15:50:55 CheckSocketConnect Entered 
DateTime 05/30/2023 15:50:55, SubscribeManager - ok. expected subscribe within threshold limit of SubscribeTimeout. No action needed 
DateTime 05/30/2023 15:51:05 CheckSocketConnect (HttpClient Or Task.Factory) Failed The request was canceled due to the configured HttpClient.Timeout of 10 seconds elapsing. 
DateTime 05/30/2023 15:51:05 ParseCheckSocketConnectException Error. The request was canceled due to the configured HttpClient.Timeout of 10 seconds elapsing. 
DateTime 05/30/2023 15:53:30 CheckSocketConnect Entered 
DateTime 05/30/2023 15:53:30, SubscribeManager - StartSubscribeHeartbeatCheckCallback - No network or no pubnub instance mapping 
DateTime: 05/30/2023 15:53:30, SendRequestAndGetJsonResponseHttpClient Exception The operation was canceled. 
DateTime 05/30/2023 15:53:30 GetTimeWithHttpClient Resp OK 
DateTime 05/30/2023 15:53:30, reconnection policy is DISABLED, please handle reconnection manually. 
DateTime 05/30/2023 15:56:05 CheckSocketConnect Entered 
DateTime 05/30/2023 15:56:05, SubscribeManager - **No auto subscribe within threshold limit of SubscribeTimeout**. Calling MultiChannelSubscribeRequest 
DateTime 05/30/2023 15:56:05, reconnection policy is DISABLED, please handle reconnection manually.  

核心错误 #

reconnection policy is DISABLED, please handle reconnection manually

可能原因与解决办法 #

• 启用自动重连策略：当前重连策略被禁用是核心问题，在PNConfiguration中开启自动重连：

  var config = new PNConfiguration(new UserId($"PubNub:UserId"))
  {
      SubscribeKey = config.GetValue<string>("PubNub:SubscribeKey"), 
      PublishKey = config.GetValue<string>("PubNub:PublishKey"),
      ReconnectionPolicy = PNReconnectionPolicy.LINEAR // 或EXPONENTIAL，按需选择
  };
  _pubNub = new Pubnub(config);
  

• 调整初始化顺序：先添加监听器再执行订阅，确保回调注册完成后再接收消息：

  _pubNub.AddListener(CreateListener());
  _pubNub.Subscribe<string>()
      .Channels(new[] { "PubNub:BroadcastChannel" })
      .Execute();
  

• 添加连接状态监听：在回调中处理连接断开、重连事件，主动干预：

  private SubscribeCallbackExt CreateListener()
  {
      return new SubscribeCallbackExt
      (
          messageCallback: async (pnObj, pubMsg) => await HandleReceivedMessage(pubMsg),
          statusCallback: (pnObj, status) =>
          {
              if (status.Category == PNStatusCategory.PNDisconnectedCategory)
              {
                  // 记录日志并尝试手动重连
                  pnObj.Subscribe<string>().Channels(new[] { "PubNub:BroadcastChannel" }).Execute();
              }
              else if (status.Category == PNStatusCategory.PNReconnectedCategory)
              {
                  // 记录重连成功日志
              }
          }
      );
  }
  

• 延长超时时间：针对日志中的HttpClient超时问题，调整PubNub配置的超时参数：

  config.SubscribeTimeout = TimeSpan.FromSeconds(30);
  config.NonSubscribeTimeout = TimeSpan.FromSeconds(15);
  

• 临时恢复方案：若无法立即部署代码修改，可设置IIS应用池定期回收（如每日凌晨），避免服务长时间运行后断开无法恢复。

内容的提问来源于stack exchange，提问作者Eternal21