.NET 6 Windows服务运行24小时后HttpClient调用持续超时求助

问题背景 #

.NET 6.0 Windows服务每隔几秒向API发送数据，运行约24小时后持续出现超时异常，仅能通过重启服务解决。已配置PooledConnectionLifetime为2分钟、自定义超时，三个API各有独立静态HttpClient实例。

现有API调用代码 #

private static HttpClient client = new HttpClient(); 

public static void SetTimeout(int timeout, int poolConnectionLifeTime)
{
    var handler = new SocketsHttpHandler
    {
        PooledConnectionLifetime = TimeSpan.FromMinutes(poolConnectionLifeTime) 
    };
    client = new HttpClient(handler);
    client.Timeout = TimeSpan.FromSeconds(timeout);
    TokenRetriever.Log.Info("Auth API http client initialised");
}      

public async Task<string> ExecuteGetAndReturnResponse(Dictionary<string, string> requestParams, string url)
{
    string responseReturned = string.Empty;

    try
    {
        using (var response = await client.PostAsync(url, new FormUrlEncodedContent(requestParams)))
        {
            if (response.IsSuccessStatusCode)
            {
                using (var stream = await response.Content.ReadAsStreamAsync())
                {
                    using (var streamReader = new StreamReader(stream))
                    {
                        responseReturned = await streamReader.ReadToEndAsync();
                    }
                }
            }
            else
            {
                responseReturned = $"ERROR: - {response.StatusCode} - {response.ReasonPhrase}";
                TokenRetriever.Log.Info($"Response returned {responseReturned}");
            }
        }
    }
    catch (TaskCanceledException ex) when (ex.InnerException is TimeoutException)
    {
       TokenRetriever.Log.Error("Fetching token Timed out: " + ex.Message);
    }
    catch (TaskCanceledException ex)
    {
        // Handle cancellation.
       TokenRetriever.Log.Error("Task Cancellation fetching token error: " + ex.Message);
    }

    return responseReturned;
}

解决方案建议 #

• 修复静态HttpClient初始化的线程安全问题：当前SetTimeout直接替换静态client实例，多线程场景下可能出现竞态条件，导致部分请求使用未正确配置的旧实例。建议用Lazy<HttpClient>实现线程安全的惰性初始化：

  private static readonly Lazy<HttpClient> _client = new Lazy<HttpClient>(() =>
  {
      var handler = new SocketsHttpHandler
      {
          PooledConnectionLifetime = TimeSpan.FromMinutes(2)
      };
      var client = new HttpClient(handler);
      client.Timeout = TimeSpan.FromSeconds(30);
      TokenRetriever.Log.Info("Auth API http client initialised");
      return client;
  });
  
  private static HttpClient Client => _client.Value;
  

• 补充连接池相关配置：除PooledConnectionLifetime外，设置PooledConnectionIdleTimeout（如1分钟）避免闲置连接长期占用；调整MaxConnectionsPerServer为合理值，防止连接池耗尽。
• 添加重试机制：针对超时、服务端错误实现指数退避重试，可使用Polly库简化逻辑，避免单次超时引发持续失败：

  var retryPolicy = Policy
      .Handle<TaskCanceledException>(ex => ex.InnerException is TimeoutException)
      .Or<HttpRequestException>()
      .WaitAndRetryAsync(3, retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)));
  
  var response = await retryPolicy.ExecuteAsync(async () => 
      await Client.PostAsync(url, new FormUrlEncodedContent(requestParams)));
  

• 处理DNS缓存问题：.NET HttpClient默认缓存DNS解析结果，若API IP变更会导致旧IP失效。在SocketsHttpHandler中配置DnsRefreshTimeout（如5分钟）强制刷新DNS：

  var handler = new SocketsHttpHandler
  {
      PooledConnectionLifetime = TimeSpan.FromMinutes(2),
      DnsRefreshTimeout = TimeSpan.FromMinutes(5)
  };
  

• 增强日志维度：记录请求ID、连接耗时、活跃连接数，超时异常时补充请求上下文信息，帮助定位是连接池、API端还是网络问题。
• 改用IHttpClientFactory管理实例：依赖注入方式的IHttpClientFactory会自动处理连接池和实例生命周期，避免手动管理静态HttpClient的潜在问题，更适合长期运行的服务场景。

内容的提问来源于stack exchange，提问作者AndrewE