• 记一次 .NET 某企业OA后端服务 卡死分析


    一:背景

    1.讲故事

    前段时间有位朋友微信找到我,说他生产机器上的 Console 服务看起来像是卡死了,也不生成日志,对方也收不到我的httpclient请求,不知道程序出现什么情况了,特来寻求帮助。

    哈哈,一般来说卡死的情况在窗体程序(WinForm,WPF) 上特别多,在 Console,WebApi 中相对较少,既然找到我,那就上 WinDbg 分析吧。

    二:WinDbg 分析

    1. 程序真的卡死了吗

    程序之所以能跑的梭梭响,全靠线程帮忙,如果玩不转可能就是线程上出了点什么问题,接下来使用 !t 展示下线程列表。

    
    0:000> !t
    ThreadCount:      124
    UnstartedThread:  0
    BackgroundThread: 105
    PendingThread:    0
    DeadThread:       18
    Hosted Runtime:   no
                                                                                                                Lock  
     DBG   ID     OSID ThreadOBJ           State GC Mode     GC Alloc Context                  Domain           Count Apt Exception
       0    1     65ec 000002097A4BF390    2a020 Preemptive  0000000000000000:0000000000000000 000002097a4bea20 -00001 MTA 
       3    2    15afc 000002097A50BA10    2b220 Preemptive  00000209060BE8F8:00000209060BEBE8 000002097a4bea20 -00001 MTA (Finalizer) 
       4    3    11d20 000002097A524EC0  102a220 Preemptive  0000000000000000:0000000000000000 000002097a4bea20 -00001 MTA (Threadpool Worker) 
       5    9     b8a4 000002097DE10290  3029220 Preemptive  0000000000000000:0000000000000000 000002097a4bea20 -00001 MTA (Threadpool Worker) 
       6   13    1b22c 000002097DE0ADB0  1029220 Preemptive  0000000000000000:0000000000000000 000002097a4bea20 -00001 MTA (Threadpool Worker) 
       7   16    1b0e8 000002097DE1A030  202b220 Preemptive  00000209063CB630:00000209063CC1F8 000002097a4bea20 -00001 MTA 
       8   17    138c8 000002097DE175C0  202b220 Preemptive  00000209063F71C0:00000209063F7BC8 000002097a4bea20 -00001 MTA 
       9   18    1afd0 000002097DE181E0  202b220 Preemptive  00000209064627E0:0000020906463598 000002097a4bea20 -00001 MTA 
      10   19    1ac48 000002097DE13310  202b220 Preemptive  000002090632D6C8:000002090632E0D8 000002097a4bea20 -00001 MTA 
      11   20    18704 000002097DE16390  202b220 Preemptive  00000209063FB5A8:00000209063FBBC8 000002097a4bea20 -00001 MTA 
      12   21    1ade4 000002097DE187F0  202b220 Preemptive  00000209062EA138:00000209062EA708 000002097a4bea20 -00001 MTA 
      13   22    13164 000002097DE13920  202b220 Preemptive  0000020906392108:0000020906392A38 000002097a4bea20 -00001 MTA 
      14   23    1b334 000002097DE169A0  202b220 Preemptive  00000209063CD9E8:00000209063CE1F8 000002097a4bea20 -00001 MTA 
      ...
     106  168    19e18 0000020927066770  a029220 Preemptive  0000000000000000:0000000000000000 000002097a4bea20 -00001 MTA (Threadpool Completion Port) 
     108  136     af74 0000020928590290  8029220 Preemptive  00000209063B60E0:00000209063B6408 000002097a4bea20 -00001 MTA (Threadpool Completion Port) 
     107   84     e7f0 000002097AD6B600  8029220 Preemptive  00000209063E3300:00000209063E3DD8 000002097a4bea20 -00001 MTA (Threadpool Completion Port) 
    
    

    一般来说卦中的 Lock Count 列表示当前线程所持有的托管锁个数,现在显示的 -00001 应该是命令不起效果了。。。不过没关系,我们还可以通过 !syncblk 来看下 lock 的情况,毕竟 95% 的锁场景都会用到它。

    
    0:000> !syncblk
    Index         SyncBlock MonitorHeld Recursion Owning Thread Info          SyncBlock Owner
      176 000002097E0DA1D8           17         1 00000209216A3500 132c8  71   00000209002c6a68 System.Object
      191 000002097E0DAC88           27         1 000002097ADC54F0 1ae90  53   00000209000185e8 System.IO.TextWriter+SyncTextWriter
    -----------------------------
    Total           270
    CCW             0
    RCW             0
    ComClassFactory 0
    Free            32
    
    

    从卦中的 MonitorHeld=27 来看,表示这个 SyncTextWriter 对象当前有 13 个线程在等待,有 1 个线程在持有,那这个线程为什么没有退出呢? 接下来可以切到 53 号线程上,查看下它的线程栈。

    
    0:053> ~~[1ae90]s
    ntdll!NtWriteFile+0x14:
    00007ffd`a70df774 c3              ret
    0:053> !clrstack 
    OS Thread Id: 0x1ae90 (53)
            Child SP               IP Call Site
    00000070505BDE88 00007ffda70df774 [InlinedCallFrame: 00000070505bde88] Interop+Kernel32.WriteFile(IntPtr, Byte*, Int32, Int32 ByRef, IntPtr)
    00000070505BDE88 00007ffd32a7cbe2 [InlinedCallFrame: 00000070505bde88] Interop+Kernel32.WriteFile(IntPtr, Byte*, Int32, Int32 ByRef, IntPtr)
    00000070505BDE50 00007ffd32a7cbe2 ILStubClass.IL_STUB_PInvoke(IntPtr, Byte*, Int32, Int32 ByRef, IntPtr)
    00000070505BDF20 00007ffd3a2d6971 System.ConsolePal+WindowsConsoleStream.WriteFileNative(IntPtr, Byte[], Int32, Int32, Boolean)
    00000070505BDF80 00007ffd3a2d672f System.ConsolePal+WindowsConsoleStream.Write(Byte[], Int32, Int32) [/_/src/libraries/System.Console/src/System/ConsolePal.Windows.cs @ 1131]
    00000070505BDFD0 00007ffd377c05d9 System.IO.StreamWriter.Flush(Boolean, Boolean) [/_/src/libraries/System.Private.CoreLib/src/System/IO/StreamWriter.cs @ 260]
    00000070505BE050 00007ffd3a2d6687 System.IO.StreamWriter.WriteLine(System.String)
    00000070505BE0D0 00007ffd3a2d6472 System.IO.TextWriter+SyncTextWriter.WriteLine(System.String) [/_/src/libraries/System.Private.CoreLib/src/System/IO/TextWriter.cs @ 880]
    00000070505BE130 00007ffd3a2d640d System.Console.WriteLine(System.String) [/_/src/libraries/System.Console/src/System/Console.cs @ 716]
    00000070505BE160 00007ffd3a2375ab HCloud.xxx+d__2.MoveNext()
    00000070505BEAE0 00007ffd37e19365 System.Runtime.CompilerServices.AsyncMethodBuilderCore.Start[[System.__Canon, System.Private.CoreLib]](System.__Canon ByRef) [/_/src/libraries/System.Private.CoreLib/src/System/Runtime/CompilerServices/AsyncMethodBuilderCore.cs @ 63]
    00000070505BEB40 00007ffd3a235cc3 HCloud.xxxx.Execute(Quartz.IJobExecutionContext)
    00000070505BEBB0 00007ffd3a231264 Quartz.xxx+d__9.MoveNext()
    00000070505BEE90 00007ffd3a230183 System.Runtime.CompilerServices.AsyncMethodBuilderCore.Start[[Quartz.Core.JobRunShell+d__9, Quartz]](d__9 ByRef) [/_/src/libraries/System.Private.CoreLib/src/System/Runtime/CompilerServices/AsyncMethodBuilderCore.cs @ 63]
    00000070505BEF10 00007ffd3a2300db System.Runtime.CompilerServices.AsyncTaskMethodBuilder.Start[[Quartz.Core.JobRunShell+d__9, Quartz]](d__9 ByRef) [/_/src/libraries/System.Private.CoreLib/src/System/Runtime/CompilerServices/AsyncTaskMethodBuilder.cs @ 33]
    00000070505BEF40 00007ffd3a230082 Quartz.xxxxl.Run(System.Threading.CancellationToken)
    00000070505BF020 00007ffd3a22ffc3 Quartz.Core.QuartzSchedulerThread+c__DisplayClass28_0.b__0()
    00000070505BF060 00007ffd3939b71a System.Threading.Tasks.Task`1[[System.__Canon, System.Private.CoreLib]].InnerInvoke() [/_/src/libraries/System.Private.CoreLib/src/System/Threading/Tasks/Future.cs @ 507]
    00000070505BF0C0 00007ffd37d54431 System.Threading.ExecutionContext.RunFromThreadPoolDispatchLoop(System.Threading.Thread, System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object) [/_/src/libraries/System.Private.CoreLib/src/System/Threading/ExecutionContext.cs @ 300]
    00000070505BF110 00007ffd37d53657 System.Threading.Tasks.Task.ExecuteWithThreadLocal(System.Threading.Tasks.Task ByRef, System.Threading.Thread) [/_/src/libraries/System.Private.CoreLib/src/System/Threading/Tasks/Task.cs @ 2352]
    00000070505BF1C0 00007ffd37d50e04 System.Threading.ThreadPoolWorkQueue.Dispatch() [/_/src/libraries/System.Private.CoreLib/src/System/Threading/ThreadPool.cs @ 677]
    00000070505BF690 00007ffd925ea7a3 [DebuggerU2MCatchHandlerFrame: 00000070505bf690] 
    
    

    仔细观察线程栈会很惊讶的发现,居然还能卡在 System.Console.WriteLine 方法上,挺奇怪的,为了探究原因,我们使用 k 命令看下非托管栈。

    
    0:053> k 5
     # Child-SP          RetAddr               Call Site
    00 00000070`505bddd8 00007ffd`a32febda     ntdll!NtWriteFile+0x14
    01 00000070`505bdde0 00007ffd`32a7cbe2     KERNELBASE!WriteFile+0x7a
    02 00000070`505bde50 00007ffd`3a2d6971     0x00007ffd`32a7cbe2
    03 00000070`505bdf20 00007ffd`3a2d672f     System_Console!System.ConsolePal.WindowsConsoleStream.WriteFileNative+0x61
    04 00000070`505bdf80 00007ffd`377c05d9     System_Console!System.ConsolePal.WindowsConsoleStream.Write+0x3f
    
    0:053> ub ntdll!NtWriteFile+0x14
    ntdll!NtDeviceIoControlFile+0x15:
    00007ffd`a70df755 cd2e            int     2Eh
    00007ffd`a70df757 c3              ret
    00007ffd`a70df758 0f1f840000000000 nop     dword ptr [rax+rax]
    ntdll!NtWriteFile:
    00007ffd`a70df760 4c8bd1          mov     r10,rcx
    00007ffd`a70df763 b808000000      mov     eax,8
    00007ffd`a70df768 f604250803fe7f01 test    byte ptr [SharedUserData+0x308 (00000000`7ffe0308)],1
    00007ffd`a70df770 7503            jne     ntdll!NtWriteFile+0x15 (00007ffd`a70df775)
    00007ffd`a70df772 0f05            syscall
    
    

    从上面的 syscall 系统调用关键词看,代码是将用户态的 ntdll!NtWriteFile 切到入了内核态的 nt!NtWriteFile 方法,那进入了内核态为什么没有返回呢? 这又是一个值得思索的问题。

    2. 为什么进入了内核态无返回

    其实 ntdll!NtWriteFile 这个 win32 api 方法的第一个参数是一个 handle 的文件句柄,签名如下。

    
    __kernel_entry NTSYSCALLAPI NTSTATUS NtWriteFile(
      [in]           HANDLE           FileHandle,
      [in, optional] HANDLE           Event,
      [in, optional] PIO_APC_ROUTINE  ApcRoutine,
      [in, optional] PVOID            ApcContext,
      [out]          PIO_STATUS_BLOCK IoStatusBlock,
      [in]           PVOID            Buffer,
      [in]           ULONG            Length,
      [in, optional] PLARGE_INTEGER   ByteOffset,
      [in, optional] PULONG           Key
    );
    
    

    可能 handle 在内核中被别人占用了,可以用 !handle 查看下 rcx 寄存器。

    
    0:053> r
    rax=0000000000000008 rbx=00000070505bdf50 rcx=0000000000000418
    rdx=0000000000000000 rsi=0000000000000000 rdi=0000000000000418
    rip=00007ffda70df774 rsp=00000070505bddd8 rbp=00000070505bdf10
     r8=0000000002000805  r9=0000000000000000 r10=0000020906191620
    r11=00000070505bc8f8 r12=0000000000000100 r13=0000000000000053
    r14=0000000000000077 r15=000002097adc54f0
    iopl=0         nv up ei pl zr na po nc
    cs=0033  ss=002b  ds=002b  es=002b  fs=0053  gs=002b             efl=00000246
    ntdll!NtWriteFile+0x14:
    00007ffd`a70df774 c3              ret
    0:053> !handle 0000000000000418 f
    Handle 0000000000000418
      Type         	File
      Attributes   	0
      GrantedAccess	0x120196:
             ReadControl,Synch
             Write/Add,Append/SubDir/CreatePipe,WriteEA,ReadAttr,WriteAttr
      HandleCount  	3
      PointerCount 	65483
      No object specific information available
    
    

    哈哈,其实也看不出什么,也没法进入内核态,所以下一步只能到网上搜搜看,其实有经验的朋友肯定猜出来了,应该是控制台启用了 快捷编辑窗口 ,截图如下:

    快捷编辑窗口 简而言之就是用户可以在控制台上独占这个窗口,编辑一些内容, 可一旦被用户独占,那程序侧就没法输出内容到 控制台窗口 上了,只能在 内核态 傻傻等等,这应该就是形成原因,画个图大概就像下面这样。

    将信息告诉朋友后,朋友说他用的是 Windows 服务部署,但不管是什么模式部署,注释掉 Console.WriteLine 肯定没错。

    感谢如家的 邓工 提供的禁用 快捷编辑窗口 代码。

    
        #region 关闭控制台 快速编辑模式、插入模式
        const int STD_INPUT_HANDLE = -10;
        const uint ENABLE_QUICK_EDIT_MODE = 0x0040;
        const uint ENABLE_INSERT_MODE = 0x0020;
        [DllImport("kernel32.dll", SetLastError = true)]
        internal static extern IntPtr GetStdHandle(int hConsoleHandle);
        [DllImport("kernel32.dll", SetLastError = true)]
        internal static extern bool GetConsoleMode(IntPtr hConsoleHandle, out uint mode);
        [DllImport("kernel32.dll", SetLastError = true)]
        internal static extern bool SetConsoleMode(IntPtr hConsoleHandle, uint mode);
    
        public static void DisbleQuickEditMode()
        {
            IntPtr hStdin = GetStdHandle(STD_INPUT_HANDLE);
            uint mode;
            GetConsoleMode(hStdin, out mode);
            mode &= ~ENABLE_QUICK_EDIT_MODE;//移除快速编辑模式
            mode &= ~ENABLE_INSERT_MODE;      //移除插入模式
            SetConsoleMode(hStdin, mode);
        }
        #endregion
    
    

    三:总结

    这次卡死的事故,主要还是开发人员大量的使用 Console.WriteLine 来输出日志,在某个时刻输出端窗口因为各种原因被他人独占,导致程序侧无法输出内容到窗口而一直被迫等待,之后朋友将日志输出切换到文件模式,问题得以解决。

    其实这个问题很多新手朋友都会犯,特此记录下来。

    图片名称
  • 相关阅读:
    DMA
    [计网笔记] 应用层
    为博客添加网易云音乐播放器外链
    acm对拍程序 以及sublime text3的文件自动更新插件auto refresh
    操作系统---内存管理(上) 概念 覆盖交换技术 连续分配管理方式
    操作系统---死锁的检测和解除
    C++ bitset的使用
    Codeforces Round #647 (Div. 2)
    sublime Text3 实现2:1:1三分屏效果
    操作系统---信号量以及四个进程同步问题
  • 原文地址:https://www.cnblogs.com/huangxincheng/p/16790444.html
Copyright © 2020-2023  润新知