在 C# 中避免 Azure DocumentDB 中的任务取消异常

Question

我们正在运行一个 Web 应用程序，它有多个线程通过单例类访问 Azure 的 DocumentDB API，以读取或写入数据到我们的 Cosmos 数据库。在这个类中，我们在静态构造函数中初始化我们的 DocumentClient。在整个程序中，多个线程可以访问数据库，调用如下方法（假设 Uri 是一个全局定义的属性）：

    public static async Task SaveItem<T>(T item)
    {
        try
        {
            await Client.UpsertDocumentAsync(Uri, item);
        }
        catch (Exception)
        { 
          // Some Basic Exception Handling.
        }
     }

几乎每个方法都是异步的，我们从不对这些方法的调用者调用 .Wait 或 .Result 。问题是，从多个线程中保存大量项目后，我们最终会得到以下异常：

System.Threading.Tasks.TaskCanceledException: A task was canceled.

这个异常特别奇怪的是，一旦第一次抛出，任何数据库调用（读取或写入）都不会成功，都会抛出同样的 TaskCancelledException。我们找到的唯一补救办法是重新启动程序。

抛出异常时的完整堆栈跟踪打印输出如下：

System.Threading.Tasks.TaskCanceledException: A task was canceled.
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.GatewayStoreModel.<>c__DisplayClass10.<<InvokeAsync>b__f>d__12.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<>c__DisplayClass2.<<ExecuteAsync>b__0>d__4.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<ExecuteRetry>d__1b.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<ExecuteRetry>d__1b.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<ExecuteAsync>d__a.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.GatewayStoreModel.<InvokeAsync>d__1f.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.GatewayStoreModel.<ProcessMessageAsync>d__2.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.Client.DocumentClient.<ReadAsync>d__30c.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.Client.DocumentClient.<ReadDocumentPrivateAsync>d__18d.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<>c__DisplayClass2.<<ExecuteAsync>b__0>d__4.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<ExecuteRetry>d__1b.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<ExecuteRetry>d__1b.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at Microsoft.Azure.Documents.BackoffRetryUtility`1.<ExecuteAsync>d__a.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at System.Runtime.CompilerServices.ConfiguredTaskAwaitable`1.ConfiguredTaskAwaiter.GetResult()
   at AutomatedRegressionTestSuite.Managers.DatabaseManager.<GetItemAsync>d__20`1.MoveNext() in C:\Users\Administrator\Perforce\WinBuildServer\tools\Asriel\AutomatedRegressionTestSuite\Managers\DatabaseManager.cs:line 451
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at System.Runtime.CompilerServices.TaskAwaiter`1.GetResult()
   at AutomatedRegressionTestSuite.Managers.DatabaseManager.<GetItemById>d__8`1.MoveNext() in C:\Users\Administrator\Perforce\WinBuildServer\tools\Asriel\AutomatedRegressionTestSuite\Managers\DatabaseManager.cs:line 104
   --- End of inner exception stack trace ---

我们如何防止这些任务取消异常的发生，或者至少以一种方式处理它们，使得我们的整个数据库在抛出后仍然可以访问？

Answer 1

TaskCanceledException 来自 httpclient，这意味着它是一个 http 超时异常。请检查客户端的负载，负载上的异常短尖峰可能会导致此问题。否则，另一种选择是在 ConnectionPolicy 中提高 requestTimeout 以提供更多时间。