I’ve been having this issue for a while now and was never really able to resolve it. This issue only appears when I’m using the SendAsync/ReceiveAsync methods rather than the Begin/EndSend socket methods for asynchronous operations.
If have a fairly sophisticated tcp socket library and been meaning to replace the BeginSend methods with SendAsync but because of the issue I’m experiencing I always had to put it off. My socket server is handling heavy stress scenarios with > 1000 clients connected pushing constantly over 100mbit/sec and I’d like to utilize SendAsync method so I don’t have the IAsyncResult allocation overhead.
Anyways, what happens is, everything works fine as long as I’m just sending/receiving data, however under high stress scenarios when the server is trying to disconnect/shutdown a client I’m occasionally getting the following Exception:
System.InvalidOperationException was unhandled
Message=Cannot apply a context that has been marshaled across AppDomains, that was not acquired through a Capture operation or that has already been the argument to a Set call.
Source=mscorlib
StackTrace:
at System.Threading.ExecutionContext.Run(ExecutionContext executionContext, ContextCallback callback, Object state)
at System.Net.Sockets.SocketAsyncEventArgs.FinishOperationSuccess(SocketError socketError, Int32 bytesTransferred, SocketFlags flags)
at System.Net.Sockets.SocketAsyncEventArgs.CompletionPortCallback(UInt32 errorCode, UInt32 numBytes, NativeOverlapped* nativeOverlapped)
at System.Threading._IOCompletionCallback.PerformIOCompletionCallback(UInt32 errorCode, UInt32 numBytes, NativeOverlapped* pOVERLAP)
InnerException:
I’m unable to catch this exception anywhere as it seems to happen within the .NET framework and I can’t do anything about it crashing my server.
As this only happens when I call my shutdown procedure, I’m guessing it has something to do with calling Shutdown on the socket while it still has a Read/Write procedure pending.
However, I also tried to delay calling shutdown until all read/write Send/ReceiveAsync calls have returned and call shutdown after that but that didn’t help either.
Here’s how I try to shutdown sockets:
private void InternalDisconnect(SocketError socketError)
{
lock (shutdownLock)
{
if (isShutdown)
return;
isShutdown = true;
}
allowSend = false;
SocketError = socketError;
ThreadPool.QueueUserWorkItem(delegate
{
lock (padLock)
{
try
{
if (TcpSocketStatus == TcpSocketStatus.Disconnected)
return;
TcpSocketStatus = TcpSocketStatus.Disconnecting;
if (asyncSendArgs != null)
{
asyncSendArgs.Completed -= SendCompleted;
asyncSendArgs.SetBuffer(null, 0, 0);
asyncSendArgs.Dispose();
}
if (asyncReceiveArgs != null)
{
asyncReceiveArgs.Completed -= ReceiveCompleted;
asyncReceiveArgs.SetBuffer(null, 0, 0);
asyncReceiveArgs.Dispose();
}
try
{
bufferedSender.Clear();
Socket.Shutdown(SocketShutdown.Both);
if (Socket.Connected)
{
Socket.Disconnect(true);
}
}
catch
{
}
try
{
Socket.Close();
}
catch
{
}
TcpSocketStatus = TcpSocketStatus.Disconnected;
if (socketError != SocketError.Success)
{
if (log.IsDebugEnabled)
log.Debug("SocketDisconnected\tSocketError:{0}", socketError);
}
else
{
if (log.IsDebugEnabled)
log.Debug("SocketDisconnected");
}
DisconnectTime = DateTime.UtcNow;
if (TcpSocketDisconnected != null)
TcpSocketDisconnected(this);
}
catch (Exception ex)
{
log.ErrorException("InternalDisconnect", ex);
}
}
});
}
Ok, for some reason it always helps me to ask question because I tend to figure out things always shortly after :/
It seems that asyncSendArgs.SetBuffer(null, 0, 0); is the culprit.
Even though that another piece of the code should make sure that there are no outstanding async operations before InternalDisconnect is called it seems like there still are and altering the buffer while it’s in use would definitely cause an issue.
I’ve been running heavy stress tests for 15 minutes so far and everything seems fine (before I always got that exception within one minute). I’ll let it run for some more time and hope that that did it.