You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@logging.apache.org by "Emmo Emminghaus (Jira)" <ji...@apache.org> on 2019/11/14 19:34:00 UTC

[jira] [Commented] (LOG4NET-583) RolingFileAppender+InterProcessLock is locked after rolling log files

    [ https://issues.apache.org/jira/browse/LOG4NET-583?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16974560#comment-16974560 ] 

Emmo Emminghaus commented on LOG4NET-583:
-----------------------------------------

Same Experience... same result:

Deadlock in FileAppender.FileOpen().

My suggestion:

Use a different name for the {color:#172b4d}File Rolling{color} mutex...

Then you have the same behavior when using the MinimalLock… IT WORKS PERFECT!

 

PS: I hope such a simple solution does not need another 2 years.

> RolingFileAppender+InterProcessLock is locked after rolling log files
> ---------------------------------------------------------------------
>
>                 Key: LOG4NET-583
>                 URL: https://issues.apache.org/jira/browse/LOG4NET-583
>             Project: Log4net
>          Issue Type: Bug
>          Components: Appenders
>    Affects Versions: 2.0.8
>         Environment: Windows Server 2008 R2 Standart, IIS 7.5
>            Reporter: Eugene Panov
>            Priority: Major
>
> We are using log4net for logging http requests to HttpHandlers.
> RollingFileAppender is locked after rolling log files by date or by size.
> Here is the dump of locked thread:
> [[HelperMethodFrame_1OBJ] (System.Threading.WaitHandle.WaitOneNative)] System.Threading.WaitHandle.WaitOneNative(Microsoft.Win32.SafeHandles.SafeWaitHandle, UInt32, Boolean, Boolean)
> mscorlib_ni!System.Threading.WaitHandle.WaitOne(Int64, Boolean)+23
> log4net.Appender.RollingFileAppender.AdjustFileBeforeAppend()+47
> log4net.Appender.RollingFileAppender.Append(log4net.Core.LoggingEvent)+18
> log4net.Appender.AppenderSkeleton.DoAppend(log4net.Core.LoggingEvent)+c5
> log4net.Util.AppenderAttachedImpl.AppendLoopOnAppenders(log4net.Core.LoggingEvent)+9a
> log4net.Repository.Hierarchy.Logger.CallAppenders(log4net.Core.LoggingEvent)+8a
> log4net.Repository.Hierarchy.Logger.Log(System.Type, log4net.Core.Level, System.Object, System.Exception)+6f
> log4net.Core.LogImpl.ErrorFormat(System.String, System.Object, System.Object, System.Object)+10e
> FileManager.SaveReceivedDeviceLog(System.Web.HttpPostedFile, System.String)+1e1
> upload_log_file.ProcessRequest(System.Web.HttpContext)+110
> System_Web_ni!System.Web.HttpApplication+CallHandlerExecutionStep.System.Web.HttpApplication.IExecutionStep.Execute()+257
> System_Web_ni!System.Web.HttpApplication.ExecuteStep(IExecutionStep, Boolean ByRef)+ab
> System_Web_ni!System.Web.HttpApplication+PipelineStepManager.ResumeSteps(System.Exception)+511
> System_Web_ni!System.Web.HttpApplication.BeginProcessRequestNotification(System.Web.HttpContext, System.AsyncCallback)+72
> System_Web_ni!System.Web.HttpRuntime.ProcessRequestNotificationPrivate(System.Web.Hosting.IIS7WorkerRequest, System.Web.HttpContext)+269
> System_Web_ni!System.Web.Hosting.PipelineRuntime.ProcessRequestNotificationHelper(IntPtr, IntPtr, IntPtr, Int32)+411
> System_Web_ni!System.Web.Hosting.PipelineRuntime.ProcessRequestNotification(IntPtr, IntPtr, IntPtr, Int32)+2b
> System_Web_ni!DomainNeutralILStubClass.IL_STUB(Int64, Int64, Int64, Int32)+24
> System_Web_ni!DomainNeutralILStubClass.IL_STUB(IntPtr, System.Web.RequestNotificationStatus ByRef)+4b
> [[NDirectMethodFrameStandalone] (System.Web.Hosting.UnsafeIISMethods.MgdIndicateCompletion)] System.Web.Hosting.UnsafeIISMethods.MgdIndicateCompletion(IntPtr, System.Web.RequestNotificationStatusByRef)
> System_Web_ni!System.Web.Hosting.PipelineRuntime.ProcessRequestNotificationHelper(IntPtr, IntPtr, IntPtr, Int32)+540
> System_Web_ni!System.Web.Hosting.PipelineRuntime.ProcessRequestNotification(IntPtr, IntPtr, IntPtr, Int32)+2b
> System_Web_ni!DomainNeutralILStubClass.IL_STUB(Int64, Int64, Int64, Int32)+24
> I found that there are two places with mutex instanciation in the code:
> *In FileAppender.cs:*
> public override void ActivateOptions()
> 			{
> 				if (m_mutex == null)
> 				{
> 					string mutexFriendlyFilename = CurrentAppender.File
> 							.Replace("\\", "_")
> 							.Replace(":", "_")
> 							.Replace("/", "_");
> 					{color:red}m_mutex = new Mutex(false, mutexFriendlyFilename);
> {color}				}
> 				else
> 				{
> 					CurrentAppender.ErrorHandler.Error("Programming error, mutex already initialized!");
> 				}
> 			}
> and
> *in RollingFileAppender.cs:*
> // initialize the mutex that is used to lock rolling
> {color:red}m_mutexForRolling = new Mutex(false, m_baseFileName.Replace("\\", "_").Replace(":", "_").Replace("/", "_") + "_MutexForRolling");{color}
> It looks like these mutexes are used for differnt purposes but they have the same name.
> Suppose they should have different names.
> But really locking is caused by the following code:
> protected override void WriteHeader()
> 		{
> 			if (m_stream != null)
> 			{
> 				if (m_stream.AcquireLock())
> 				{
> 					try
> 					{
> 						base.WriteHeader();
> 					}
> 					finally
> 					{
> 						m_stream.ReleaseLock();
> 					}
> 				}
> 			}
> 		}
> public bool AcquireLock()
> 			{
> 				bool ret = false;
> 				lock (this)
> 				{
> 					if (m_lockLevel == 0)
> 					{
> 						// If lock is already acquired, nop
> 						m_realStream = m_lockingModel.AcquireLock();
> 					}
> 					if (m_realStream != null)
> 					{
> 						m_lockLevel++;
> 						ret = true;
> 					}
> 				}
> 				return ret;
> 			}
> m_realStream = m_lockingModel.AcquireLock(); - this line executes th following function:
> public override Stream AcquireLock()
> 			{
> 				if (m_mutex != null)
>                 {
> 					// TODO: add timeout?
> 					m_mutex.WaitOne();
> 					// increment recursive watch
> 					m_recursiveWatch++;
>                     
> 					// should always be true (and fast) for FileStream
>                     if (m_stream != null)
>                     {
> 					
> 						if (m_stream.CanSeek)
> 						{
> 							m_stream.Seek(0, SeekOrigin.End);
> 						}
> 					}
> 					else
> 					{
> 						// this can happen when the file appender cannot open a file for writing
> 					}
> 				}
> 				else
> 				{
> 					CurrentAppender.ErrorHandler.Error("Programming error, no mutex available to acquire lock! From here on things will be dangerous!");
> 				}
> 				return m_stream;
> 			}
> if m_stream  == null then m_mutex.WaitOne(); is executed
> but m_lockLevel is not incremented
> if (m_realStream != null)
> 					{
> 						m_lockLevel++;
> 						ret = true;
> 					}
> And mutex will not be released:
> public void ReleaseLock()
> 			{
> 				lock (this)
> 				{
> 					m_lockLevel--;
> 					if (m_lockLevel == 0)
> 					{
> 						// If already unlocked, nop
> 						m_lockingModel.ReleaseLock();
> 						m_realStream = null;
> 					}
> 				}
> 			}
> It cause locking of any other threads that try to append to the log.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)