2013-05-08 90 views
1

我一直在使用Task.Factory.FromAsync()方法,并且一直在经历严重的内存泄漏。我用探查它表明,大量的对象似乎只是使用后游逛:使用Task.Factory.FromAsync时泄漏

Heap shot 140 at 98.591 secs: size: 220177584, object count: 2803125, class count: 98, roots: 666 
     Bytes  Count Average Class name 
     25049168  142325  175 System.Threading.Tasks.Task<System.Int32> (bytes: +398816, count: +2266) 
      1 root references (1 pinning) 
      142324 references from: System.Threading.Tasks.Task 
      142305 references from: System.Threading.Tasks.TaskCompletionSource<System.Int32> 
      98309 references from: task_test.Task3Test.<Run>c__AnonStorey1 
     25049024  142324  176 System.Threading.Tasks.Task (bytes: +398816, count: +2266) 
      142304 references from: System.Threading.Tasks.TaskContinuation 
     17078880  142324  120 System.Action<System.Threading.Tasks.Task<System.Int32>> (bytes: +271920, count: +2266) 
      142324 references from: System.Threading.Tasks.TaskActionInvoker.ActionTaskInvoke<System.Int32> 
     17076600  142305  120 System.Runtime.Remoting.Messaging.MonoMethodMessage (bytes: +271680, count: +2264) 
      1 root references (1 pinning) 
      142304 references from: System.MonoAsyncCall 
     17076584  142305  119 System.AsyncCallback (bytes: +271920, count: +2266) 
      1 root references (1 pinning) 
      142304 references from: System.MonoAsyncCall 
     17076584  142305  119 System.Func<System.Int32> (bytes: +271920, count: +2266) 
      1 root references (1 pinning) 
      142305 references from: System.Func<System.IAsyncResult,System.Int32> 
      142304 references from: System.Runtime.Remoting.Messaging.AsyncResult 
      1 references from: System.Func<System.AsyncCallback,System.Object,System.IAsyncResult> 
     17076584  142305  119 System.Func<System.IAsyncResult,System.Int32> (bytes: +271920, count: +2266) 
      1 root references (1 pinning) 
      142305 references from: System.Threading.Tasks.TaskFactory.<FromAsyncBeginEnd>c__AnonStorey3A<System.Int32> 
     17076480  142304  120 System.Runtime.Remoting.Messaging.AsyncResult (bytes: +271800, count: +2265) 
      98461 references from: System.Object[] 

我试图找出什么类型的东西可能/可能不会出现阻止了从识别物体的gc不再被使用。 FromAsync返回一个从TaskCompletionSource获得的Task对象,该对象具有一个类变量“source”,该变量包含Task从任务调用中获取的值。

这里是测试用例。它还包括使用StartNew()的情况,其中内存使用没有爆炸。下面的初始Test3Task没有使用ContinueWith,但是看看它是不是我们没有清理的东西,我们把它放在了(没有效果)。 [不,下面使用的聆听变量是多余的 - 有计划,使测试更加聪明,但一做永远是一样好。]

using System; 
using System.Threading; 
using System.Threading.Tasks; 

namespace task_test 
{ 
    class MainClass 
    { 
      public static void Main (string[] args) 
      { 
        // Test3 - Leaky 
        var t = new Task3Test(); 

        // Test4 - Doesn't leak 
        // var t = new Task4Test(); 

        t.Run(); 

      } 
    } 

    public class BaseTask 
    { 
      public int GetRandomInt(int top) 
      { 
        Random random = new Random(); 

        return random.Next(1,top); 
      } 
    } 

    public class FibArgs 
    { 
      public byte[] data; 
      public int n; 
    } 

    public class Fib 
    { 
      public int Calculate(FibArgs args) 
      { 
        int n = args.n; 

        int a = 0; 
        int b = 1; 
        // In N steps compute Fibonacci sequence iteratively. 
        for (int i = 0; i < n; i++) 
        { 
          int temp = a; 
          a = b; 
          b = temp + b; 
        } 
        Console.WriteLine("ThreadId: {2}, fib({0}) = {1}", n, a, Thread.CurrentThread.GetHashCode()); 
        return a; 
      } 
    } 

    public class Task3Test : BaseTask 
    { 
      public void Run() 
      { 
        bool listening = true; 
        long i = 0; 
        while (listening) 
        { 
          i++; 

          Func<int> fun =() => { 
            int n = GetRandomInt(100); 
            Fib f = new Fib(); 
            FibArgs args = new FibArgs(); 
            args.n = n; 

            return f.Calculate(args); 
          }; 

          var t = Task<int>.Factory.FromAsync(fun.BeginInvoke, fun.EndInvoke, null); 
          t.ContinueWith(x => { 
                if (x.IsCompleted) { 
                  x.Dispose(); 
                  x = null; 
                } 
              } 
            ); 
        } 
      } 
    } 

    public class Task4Test : BaseTask 
    { 
      public void Run() 
      { 
        bool listening = true; 
        long i = 0; 
        while (listening) 
        { 
          int n = GetRandomInt(100); 
          Fib f = new Fib(); 
          FibArgs args = new FibArgs(); 
          args.n = n; 

          Task.Factory.StartNew(() => f.Calculate(args), TaskCreationOptions.LongRunning) 
            .ContinueWith(x => { 
              if(x.IsFaulted) 
              { 
                Console.WriteLine("OOPS, error!!!"); 
                x.Exception.Handle(_ => true); //just an example, you'll want to handle properly 

              } 
              else if(x.IsCompleted) 
              { 
                Console.WriteLine("Cleaning up task {0}", x.Id); 
                x.Dispose(); 
              } 
            } 
          ); 
        } 
      } 

    } 
} 
+0

我在Darwin/x86上使用SGen运行此操作,几分钟后内存使用情况非常稳定。然而,Boehm,它迅速升高。你在哪个平台上使用哪种GC? – 2013-05-08 17:10:37

+0

我的同事和我在Ubuntu 12.04 64位上测试了类似的东西,并且看到了与GC相似的行为。 – 2013-05-08 18:55:09

+0

同时运行。使用sgen获得堆积,并且它在瞬间完成记忆。 – 2013-05-08 18:59:10

回答

0

我敢肯定,问题是你创建任务比完成更快,所以他们排队。

+0

当我们每秒钟敲击几次时,我们可以看到这一点与我们的http侦听器包装。 – 2013-05-09 14:29:06

+0

你确定你的任务能够快速完成,不会排队吗?你能修改你的测试用例,以便它能够证明你实际看到的问题吗? – 2013-05-14 17:42:37

+0

我现在可以肯定地确认这个特定的测试用例由于任务排队而泄漏内存。 – 2013-05-14 18:25:33