为什么.NET 3.5 Garbage Collector冻结了我的应用程序?

时间:2018-11-02 22:52:17

标签: .net windows garbage-collection windbg freeze

至少有二十个雇主的客户报告说,我们的.NET 3.5应用程序将挂起/冻结/停止响应,通常持续10到90秒,然后再次响应。

我们无法重现该问题。我们最终能够让一位客户发送在冻结应用程序时通过任务管理器创建的一些内存转储。

我很惊讶地发现.NET 3.5 Garbage Collector似乎正在暂停该应用程序。这是一些WinDbg输出(!load wow64exts!sw的存在,因为它是在64位Windows上运行的32位应用程序,并且内存转储是使用Task的64位版本创建的管理器,唯一可以使用WinDbg SOS命令的方法就是使用wow64exts。

可能是什么原因?

我必须解决或解决哪些问题?

转储文件在此处(请注意-它是1.5GB): https://1drv.ms/u/s!AolCnuRwlhMVhBi_F2UVBTYkoVtY

在主GUI线程上输出k!clrstack

0:000> !load wow64exts
0:000> !sw
Switched to Guest (WoW) mode
0:000:x86> .loadby sos mscorwks
0:000:x86> k
 # ChildEBP RetAddr  
00 01cfe97c 760b45c9 ntdll_77100000!NtWaitForSingleObject+0xc
01 01cfe9f0 710858fd KERNELBASE!WaitForSingleObjectEx+0x99
02 01cfea34 71085849 mscorwks!CLREventWaitHelper+0x2f
03 01cfea84 71085868 mscorwks!CLREvent::WaitEx+0x117
04 01cfea98 7114eaa5 mscorwks!CLREvent::Wait+0x17
05 01cfeaac 711b0f70 mscorwks!WKS::GCHeap::WaitUntilGCComplete+0x34
06 01cfeae8 710fb420 mscorwks!Thread::RareDisablePreemptiveGC+0x1b4
07 01cfeb84 6f16a19f mscorwks!JIT_RareDisableHelper+0x16
08 01cfec18 6f169c77 System_Windows_Forms_ni+0x20a19f
09 01cfec70 6f169ac1 System_Windows_Forms_ni+0x209c77
0a 01cfeca0 6f67fd12 System_Windows_Forms_ni+0x209ac1
0b 01cfecb4 0b8e9c91 System_Windows_Forms_ni+0x71fd12
WARNING: Frame IP not in any known module. Following frames may be wrong.
0c 01cfece4 03c70f42 0xb8e9c91
0d 01cfed20 71081b6c 0x3c70f42
0e 01cfed30 710985ab mscorwks!CallDescrWorker+0x33
0f 01cfedb0 710a064b mscorwks!CallDescrWorkerWithHandler+0xa3
10 01cfeef4 710a067e mscorwks!MethodDesc::CallDescr+0x19c
11 01cfef10 710a069c mscorwks!MethodDesc::CallTargetWorker+0x1f
12 01cfef28 7110579c mscorwks!MethodDescCallSite::Call+0x1a
13 01cff08c 711607bd mscorwks!ClassLoader::RunMain+0x20d
14 01cff2f4 71160cda mscorwks!Assembly::ExecuteMainMethod+0xa6
15 01cff7c4 71160ec4 mscorwks!SystemDomain::ExecuteMainMethod+0x45e
16 01cff814 71160df4 mscorwks!ExecuteEXE+0x59
17 01cff85c 7163d93b mscorwks!_CorExeMain+0x15c
18 01cff89c 7177e8b9 mscoreei!_CorExeMain+0x10e
19 01cff8b0 71784e18 mscoree!ShellShim__CorExeMain+0xa9
1a 01cff8b8 75d68484 mscoree!_CorExeMain_Exported+0x8
1b 01cff8cc 7716302c kernel32!BaseThreadInitThunk+0x24
1c 01cff914 77162ffa ntdll_77100000!__RtlUserThreadStart+0x2f
1d 01cff924 00000000 ntdll_77100000!_RtlUserThreadStart+0x1b
0:000:x86> !clrstack
OS Thread Id: 0x26ec (0)
ESP       EIP     
01cfeb90 0000002b [InlinedCallFrame: 01cfeb90] System.Windows.Forms.UnsafeNativeMethods.WaitMessage()
01cfeb8c 6f16a188 System.Windows.Forms.Application+ComponentManager.System.Windows.Forms.UnsafeNativeMethods.IMsoComponentManager.FPushMessageLoop(Int32, Int32, Int32)
01cfec28 6f169c77 System.Windows.Forms.Application+ThreadContext.RunMessageLoopInner(Int32, System.Windows.Forms.ApplicationContext)
01cfec7c 6f169ac1 System.Windows.Forms.Application+ThreadContext.RunMessageLoop(Int32, System.Windows.Forms.ApplicationContext)
01cfecac 6f67fd12 System.Windows.Forms.Application.Run()
01cfecbc 0b8e9c91 NavSim.Application.StartupManager.Run()
01cfecec 03c70f42 NavSim.Application.EntryPoint.Main(System.String[])
01cfef48 71081b6c [GCFrame: 01cfef48] 

!threads的输出:

0:000:x86> !threads
ThreadCount: 30
UnstartedThread: 0
BackgroundThread: 25
PendingThread: 0
DeadThread: 1
Hosted Runtime: no
                                      PreEmptive   GC Alloc           Lock
       ID OSID ThreadOBJ    State     GC       Context       Domain   Count APT Exception
   0    1 26ec 01ef7458      6020 Enabled  00000000:00000000 01ef2e28     0 STA
   2    2  f70 01f06878      b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Finalizer)
XXXX    3    0 07a02478   1801820 Enabled  00000000:00000000 01ef2e28     0 Ukn (Threadpool Worker)
   5    4 2970 0b921530      b220 Enabled  00000000:00000000 01ef2e28     0 MTA
   8    5 2824 07a0c1b8   200b220 Enabled  00000000:00000000 01ef2e28     0 MTA
   4    9 1dac 0ba71370       220 Enabled  00000000:00000000 01ef2e28     0 Ukn
  11    a 2498 0ba04f50      b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  12    b 1ab4 0bab8f78      b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  14    d 2b44 0bab8b90      b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  17   10 128c 0bab9b30   200b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  18   11  35c 0bab7fd8      b220 Enabled  00000000:00000000 01ef2e28     5 MTA
  19   12 2fd4 27c47580      b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  15   14 3308 27c48520       220 Enabled  00000000:00000000 01ef2e28     0 STA
  22    e  b44 27c48cf0   200b020 Enabled  00000000:00000000 01ef2e28     0 MTA
  23    c 1694 0bab83c0   200b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  24   1a 19e8 281657c0    80a220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Completion Port)
  25   16  24c 28165ba8   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  26   1e  e5c 0bab9748   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  27   21 3258 28152a38   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  28    f 1248 28151e80   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  29   20  85c 28151a98   880b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Completion Port)
  30   1f 22c0 28152268   200b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  31   1b 21e0 28154978      b020 Enabled  6816529c:68166080 01ef2e28     0 MTA
  32   19 229c 28152650   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  36   17 2270 28152e20   200b220 Enabled  00000000:00000000 01ef2e28     0 MTA
  37    8 2348 28150ee0      b020 Disabled 681660a4:681660a4 01ef2e28     1 MTA (GC)
  38   13 1234 28153208   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  39   18 16b0 281535f0   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  40    7 2f6c 281539d8   180b222 Disabled 00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)
  41    6  3f4 281512c8   180b220 Enabled  00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)

!threadpool的输出:

0:000:x86> !threadpool
CPU utilization 81%
Worker Thread: Total: 9 Running: 9 Idle: 0 MaxLimit: 1000 MinLimit: 4
Work Request in Queue: 940
AsyncTimerCallbackCompletion TimerInfo@27ff74e0
AsyncTimerCallbackCompletion TimerInfo@2823e5f8
AsyncTimerCallbackCompletion TimerInfo@27ea3260
AsyncTimerCallbackCompletion TimerInfo@41444968
AsyncTimerCallbackCompletion TimerInfo@2823eb08
AsyncTimerCallbackCompletion TimerInfo@41444f08
AsyncTimerCallbackCompletion TimerInfo@27ffb938
AsyncTimerCallbackCompletion TimerInfo@27ffbb30
AsyncTimerCallbackCompletion TimerInfo@27ffb620
AsyncTimerCallbackCompletion TimerInfo@27ff7888
AsyncTimerCallbackCompletion TimerInfo@27ea4028
AsyncTimerCallbackCompletion TimerInfo@2823d6c8
AsyncTimerCallbackCompletion TimerInfo@27ea32f0
AsyncTimerCallbackCompletion TimerInfo@2823e4d8
AsyncTimerCallbackCompletion TimerInfo@07a93d18
AsyncTimerCallbackCompletion TimerInfo@2823ed48
AsyncTimerCallbackCompletion TimerInfo@4102d320
AsyncTimerCallbackCompletion TimerInfo@4102d170
[NOTE: 922 SIMILAR LINES AS ABOVE SNIPPED FOR 30000 QUESTION CHARACTER LIMIT]
--------------------------------------
Number of Timers: 73
--------------------------------------
Completion Port Thread:Total: 1 Free: 1 MaxFree: 8 CurrentLimit: 0 MaxLimit: 1000 MinLimit: 4

线程37(标记为(GC)的线程)的本地调用堆栈(k):

0:000:x86> ~37s 
ntdll_77100000!NtWaitForSingleObject+0xc:
7716a25c c20c00          ret     0Ch
0:037:x86> k
 # ChildEBP RetAddr  
00 0bbfed80 760b45c9 ntdll_77100000!NtWaitForSingleObject+0xc
01 0bbfedf4 710858fd KERNELBASE!WaitForSingleObjectEx+0x99
02 0bbfee38 71085849 mscorwks!CLREventWaitHelper+0x2f
03 0bbfee88 71085868 mscorwks!CLREvent::WaitEx+0x117
04 0bbfee9c 711b9651 mscorwks!CLREvent::Wait+0x17
05 0bbfeef4 7114c08d mscorwks!Thread::SysSuspendForGC+0x697
06 0bbfefe4 7114a3ad mscorwks!WKS::GCHeap::SuspendEE+0x194
07 0bbff00c 7114c2fc mscorwks!WKS::GCHeap::GarbageCollectGeneration+0x136
08 0bbff098 71149c43 mscorwks!WKS::gc_heap::try_allocate_more_space+0x15a
09 0bbff0ac 71149cde mscorwks!WKS::gc_heap::allocate_more_space+0x11
0a 0bbff0cc 71082341 mscorwks!WKS::GCHeap::Alloc+0x3b
0b 0bbff0e8 71089c91 mscorwks!Alloc+0x60
0c 0bbff128 71089d56 mscorwks!FastAllocateObject+0x38
0d 0bbff1cc 29fd0b51 mscorwks!JIT_NewFast+0x9e
WARNING: Frame IP not in any known module. Following frames may be wrong.
0e 0bbff1f4 29fd0acf 0x29fd0b51
0f 0bbff214 2c1107fe 0x29fd0acf
10 0bbff22c 24454a20 0x2c1107fe
11 0bbff258 2445497d 0x24454a20
12 0bbff268 2445491f 0x2445497d
13 0bbff27c 2445456d 0x2445491f
14 0bbff2dc 2445409a 0x2445456d
15 0bbff388 24453d8a 0x2445409a
16 0bbff3a8 2c1168b3 0x24453d8a
17 0bbff3b8 2c1165a5 0x2c1168b3
18 0bbff3d8 338cb7a2 0x2c1165a5
19 0bbff3f0 338cb6a9 0x338cb7a2
1a 0bbff408 338cb666 0x338cb6a9
1b 0bbff414 338cb640 0x338cb666
1c 0bbff434 022290e1 0x338cb640
1d 0bbff45c 02227a0b 0x22290e1
1e 0bbff4cc 022267ab 0x2227a0b
1f 0bbff4f8 0222641a 0x22267ab
20 0bbff514 02347746 0x222641a
21 0bbff578 0234736b 0x2347746
22 0bbff5b8 706f88b6 0x234736b
23 0bbff5c4 707022ff mscorlib_ni+0x2188b6
24 0bbff5d8 706f8834 mscorlib_ni+0x2222ff
25 0bbff5f0 71081b6c mscorlib_ni+0x218834
26 0bbff600 710985ab mscorwks!CallDescrWorker+0x33
27 0bbff680 710a064b mscorwks!CallDescrWorkerWithHandler+0xa3
28 0bbff7b8 710a067e mscorwks!MethodDesc::CallDescr+0x19c
29 0bbff7d4 710a069c mscorwks!MethodDesc::CallTargetWorker+0x1f
2a 0bbff7ec 711bd119 mscorwks!MethodDescCallSite::Call+0x1a
2b 0bbff9d4 710b965f mscorwks!ThreadNative::KickOffThread_Worker+0x192
2c 0bbff9e8 710b95fb mscorwks!ManagedThreadBase_DispatchInner+0x4f
2d 0bbffa7c 710b951c mscorwks!ManagedThreadBase_DispatchMiddle+0xb1
2e 0bbffab8 710b96ad mscorwks!ManagedThreadBase_DispatchOuter+0x6d
2f 0bbffae0 711bceea mscorwks!ManagedThreadBase_FullTransitionWithAD+0x25
30 0bbffaf8 711bcfc4 mscorwks!ManagedThreadBase::KickOff+0x13
31 0bbffb94 711a5cf9 mscorwks!ThreadNative::KickOffThread+0x26b
32 0bbffc2c 75d68484 mscorwks!Thread::intermediateThreadProc+0x49
33 0bbffc40 7716302c kernel32!BaseThreadInitThunk+0x24
34 0bbffc88 77162ffa ntdll_77100000!__RtlUserThreadStart+0x2f
35 0bbffc98 00000000 ntdll_77100000!_RtlUserThreadStart+0x1b

任何想法都会受到赞赏;谢谢!

1 个答案:

答案 0 :(得分:1)

  

至少有二十个雇主的客户报告说,我们的.NET 3.5应用程序将挂起/冻结/停止响应,通常持续10到90秒,然后再次响应。

我已经看到,在CPU使用率很高的情况下,.NET显然认为CPU的优先级高于垃圾回收。然后,在可能的最后一刻,它执行了垃圾收集,花费了很长时间。

  

我很惊讶地发现.NET 3.5垃圾收集器似乎正在暂停该应用程序。

共有三种模式,“抢先”,“并发”和“后台”垃圾回收。抢先者可以暂停该应用程序。事外,有时候线程确实处于挂起模式。似乎不再是这种情况了。

  

可能是什么原因?

0:002:x86> !threadpool
CPU utilization 81%

仅供参考:81%是a magic value,表明正在进行垃圾回收,也防止了在线程池中创建新线程。

您已将线程37确定为GC线程,并且该线程正在挂起应用程序。

可能的原因之一是终结器阻塞。但是,这里似乎并非如此。终结器线程仍在等待终结器事件:

0:002:x86> k
 # ChildEBP RetAddr  
00 062bfa1c 760c5943 ntdll_77100000!NtWaitForMultipleObjects+0xc
01 062bfbb0 760c57f8 KERNELBASE!WaitForMultipleObjectsEx+0x133
02 062bfbcc 7114a136 KERNELBASE!WaitForMultipleObjects+0x18
03 062bfbec 7114e102 mscorwks!WKS::WaitForFinalizerEvent+0x77
04 062bfc00 710b965f mscorwks!WKS::GCHeap::FinalizerThreadWorker+0x49

但是,似乎某些线程无法挂起(PreEmptive GC列):

                                   PreEmptive   GC Alloc               Lock
        ID OSID ThreadOBJ    State     GC       Context       Domain   Count APT Exception
   37    8 2348 28150ee0      b020 Disabled 681660a4:681660a4 01ef2e28     1 MTA (GC)
   40    7 2f6c 281539d8   180b222 Disabled 00000000:00000000 01ef2e28     0 MTA (Threadpool Worker)

该工作线程似乎与您已经提到的许多计时器条目有关:

0:040:x86> k
[...]
0b 0e51f5b0 710b965f mscorwks!AddTimerCallback_Worker+0x66
[...]

让我们看看:

AsyncTimerCallbackCompletion TimerInfo@4102d170
--------------------------------------
Number of Timers: 73
--------------------------------------
Completion Port Thread:Total: 1 Free: 1 MaxFree: 8 CurrentLimit: 0 MaxLimit: 1000 MinLimit: 4
0:040:x86> dp 4102d170 L5
4102d170  715e33d8 4102d320 0835739e 711381c9
4102d180  07a67be8

0:040:x86> dp 07a67be8 L2
07a67be8  00000001 01ddd1b0

0:040:x86> !do poi(01ddd1b0)
Name: System.Threading._TimerCallback
[...]

因此,您似乎正在使用System.Threading.Timer

我不确定100%是否正确执行了此操作,但这表明您在调用回调时正在停止计时器:

0:040:x86> k
 # ChildEBP RetAddr  
00 0e51f34c 706c9f4c mscorwks!CompareExchangeMP+0x8
01 0e51f370 702866d7 mscorlib_ni+0x1e9f4c
[...]
0:040:x86> !ip2md 702866d7 
MethodDesc: 6fd71474
Method Name: System.Timers.Timer.Stop()

请确保您是using the Threading Timer correctly