Re: [Quickfix-developers] SENDINGTIME ACCURACY PROBLEM

[Facchetti - K. <s.f...@kl...>]

Is it possible to set FIX message timestamp before the message is sent 
instead of queued?
I think that the problem is here:

bool FIX::SocketConnection::send 	(  	const std::string &  	
	 )  	| [private, virtual]|

  	

Implements FIX::Responder 
<http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_responder.html#a1>. 


Definition at line 63 
<http://www.quickfixengine.org/quickfix/doc/html/_socket_connection_8cpp-source.html#l00063> 
of file SocketConnection.cpp 
<http://www.quickfixengine.org/quickfix/doc/html/_socket_connection_8cpp-source.html>. 


00064 { QF_STACK_PUSH <http://www.quickfixengine.org/quickfix/doc/html/_call_stack_8h.html#a0>(SocketConnection::send)
00065 
00066   Locker l( m_mutex );
00067 
00068   m_sendQueue <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r3>.push_back( msg );
00069   processQueue <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#a7>();
00070   signal <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#a8>();
00071   return true;
00072 
00073   QF_STACK_POP <http://www.quickfixengine.org/quickfix/doc/html/_call_stack_8h.html#a1>
00074 }
      


bool FIX::SocketConnection::processQueue 	(  	
	 )  	

  	

Definition at line 76 
<http://www.quickfixengine.org/quickfix/doc/html/_socket_connection_8cpp-source.html#l00076> 
of file SocketConnection.cpp 
<http://www.quickfixengine.org/quickfix/doc/html/_socket_connection_8cpp-source.html>. 


00077 { QF_STACK_PUSH <http://www.quickfixengine.org/quickfix/doc/html/_call_stack_8h.html#a0>(SocketConnection::processQueue)
00078 
00079   Locker l( m_mutex );
00080 
00081   if( !m_sendQueue <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r3>.size() ) return true;
00082 
00083   struct timeval timeout = { 0, 0 };
00084   fd_set writeset = m_fds <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r9>;
00085   if( select( 1 + m_socket, 0, &writeset, 0, &timeout ) <= 0 )
00086     return false;
00087     
00088   const std::string& msg = m_sendQueue <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r3>.front();
00089 
00090   int result = socket_send <http://www.quickfixengine.org/quickfix/doc/html/namespace_f_i_x.html#a2911>
00091     ( m_socket <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r0>, msg.c_str() + m_sendLength <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r4>, msg.length() - m_sendLength <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r4> );
00092 
00093   if( result > 0 )
00094     m_sendLength <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r4> += result;
00095 
00096   if( m_sendLength <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r4> == msg.length() )
00097   {
00098     m_sendLength <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r4> = 0;
00099     m_sendQueue <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r3>.pop_front();
00100   }
00101 
00102   return !m_sendQueue <http://www.quickfixengine.org/quickfix/doc/html/class_f_i_x_1_1_socket_connection.html#r3>.size();
00103 
00104   QF_STACK_POP <http://www.quickfixengine.org/quickfix/doc/html/_call_stack_8h.html#a1>
00105 }
      



Rodrick Brown ha scritto:
> You will need to optimize you're code your queues are growing too 
> large find out what's causing the bottle neck, I experienced a similar 
> bug on my system many months back and was able to overcome this issue.
>
> On Tue, Jun 10, 2008 at 8:14 AM, Facchetti - Kline 
> <s.f...@kl... <mailto:s.f...@kl...>> wrote:
>
>
>     I'm running only the initiator.
>     My FIX client machine is syncronize with time server as well FIX
>     server machine.
>     The third party FIX server can't disable sending time control.
>
>     I attach my quickfix log.
>     As you can see, I send only 19th SecurityDefinitionRequest, and
>     while replies are preocessed, I receive from the server
>     "SENDINGTIME ACCURACY PROBLEM".
>
>     Thanks,
>     Stefano
>
>     Eranga Samararathne ha scritto:
>>
>>     If you are running your initiator and acceptor in two separate
>>     machine, then make sure to synchronize time.
>>     There are few other option available to resolve this issue.
>>
>>     E.g: You can set "CheckLatency=N" property in your QuickFix
>>     configuration file.
>>
>>     For more infor read the QuickFix documentation
>>     http://www.quickfixengine.org/quickfix/doc/html/configuration.html
>>
>>     Specially read about CheckLatency and MaxLatency sections.
>>
>>     Regards,
>>
>>     Eranga.
>>
>>
>>     -----Original Message-----
>>
>>
>>     If I send many messages with the initiator (I call sendToTarget
>>     method in a loop), after some time, the server sends me this
>>     error message: "SENDINGTIME ACCURACY PROBLEM".
>>     I suppose that quickfix set the message's timestamp before
>>     putting it in a queue.. If the queue is big (like in my case),
>>     there is a delay in the message delivery.
>>
>>     How can I do..?
>>
>>     Thanks,
>>     Stefano
>>