I was recently looking into a failure in a WCF client application that runs just fine with the debugger attached but fails to run stand alone.

The error message was as follows: 

Microsoft.Hpc.ServiceBroker Warning: 0 : Service net.tcp://private.TestHeadNode:9088/346/1728/_defaultEndpoint failed. Error:The server did not provide a meaningful reply; this might be caused by a contract mismatch, a premature session shutdown or an internal server error.

    DateTime=2009-03-10T14:44:26.3899654Z

Microsoft.Hpc.ServiceBroker Warning: 10000 : Request urn:uuid:729a078e-7c79-479d-af75-976a2a2b7219 from user Anonymous user has been given up because Request has failed more than 3 times, broker will not deliver it again

    DateTime=2009-03-10T14:44:28.2337272Z

 

Following my previous blog post the trace logs were obtained and analyzed using the Service Trace Viewer Tool. It was quite evident from the traces that the message size was bigger than the service host on the compute nodes were able to handle which led to the premature session.

 

By following the direction on handling large messages here: http://msdn.microsoft.com/en-us/library/cc907051(VS.85).aspx#message_size the problem was solved and the application runs as expected (with or without the debugger attached).