Problem with ServiceBroker

I use 2 SQL Server: 1 – SQL Server 2005 Enterprise on Windows Server 2003 (Itanium) and SQL Server Express on Windows XP Professional (Pentium).

I send the message from enterprise to express the message (at once I shall tell, that sending of the message in the opposite direction gives the same result) and I process the message on express. At first sight like all is correct. Messages I receive all, I read also all of them are correct. But further there are two strange things for me.

In the table sys.conversation_endpoints there is a record (one, irrespective of how many messages in one conversation) with “state=CD”. I read that they should be deleted after 30 minutes, but they did not. Using command … WITH CLEANUP helps but in a parallel way in Service Broker Profiler I observe occurrence of a mistake which occurs after closing dialogue.

I am disturbed with this mistake and the most important is that I not understand why it arises.

EXPRESS
Event Class	EventSubClass	TextData
Broker:Message Classify	1 - Local
Broker:Conversation Group	1 - Create
Broker:Conversation	12 - Dialog Created	STARTED_INBOUND
Broker:Conversation	6 - Received Sequenced Message	CONVERSING
Broker:Remote Message Acknowledgement	3 - Message with Acknowledgement Received
Broker:Conversation	6 - Received Sequenced Message	CONVERSING
Broker:Remote Message Acknowledgement	3 - Message with Acknowledgement Received
Broker:Conversation	7 - Received END CONVERSATION	DISCONNECTED_INBOUND
Broker:Remote Message Acknowledgement	3 - Message with Acknowledgement Received
Broker:Activation	1 - Start
Broker:Message Classify	2 - Remote
Broker:Remote Message Acknowledgement	2 - Acknowledgement Sent
Broker:Conversation	2 - END CONVERSATION	CLOSED
Broker:Conversation Group
Broker:Activation	2 - Ended
Broker:Connection	4 - Closing	An error occurred while receiving data: '64(error not found)'.
Broker:Connection	5 - Closed

Enterprise
Event Class	EventSubClass	TextData
Broker:Conversation Group	1 - Create
Broker:Conversation	11 - BEGIN DIALOG	STARTED_OUTBOUND
Broker:Conversation	1 - SEND Message	CONVERSING
Broker:Message Classify	2 - Remote
Broker:Conversation	1 - SEND Message	CONVERSING
Broker:Message Classify	2 - Remote
Broker:Conversation	1 - SEND Message	CONVERSING
Broker:Message Classify	2 - Remote
Broker:Conversation	2 - END CONVERSATION	DISCONNECTED_OUTBOUND
Broker:Conversation Group	2 - Drop
Broker:Message Classify	2 - Remote
Broker:Message Classify	2 - Remote
Broker:Remote Message Acknowledgement	1 - Message with Acknowledgement Sent
Broker:Remote Message Acknowledgement	1 - Message with Acknowledgement Sent
Broker:Remote Message Acknowledgement	1 - Message with Acknowledgement Sent
Broker:Message Classify	1 - Local
Broker:Conversation	10 - Received END CONVERSATION Ack	CLOSED
Broker:Remote Message Acknowledgement	4 - Acknowledgement Received
Broker:Connection	4 - Closing
Broker:Connection	5 - Closed

Yours faithfully,

From what I understand, you are doing a BEGIN DIALOG, followed by two SENDs and then END DIALOG, all from the initiator side (the enterprise instance).

The target side (express) is just passively receiving the messages, reacting with an END CONVERSATION itself when the it receives the END CONVERSATION message from the initiator.

In general, this pattern of BEGIN DIALOG/SEND/END (aka 'fire and forget') is not advisable. If the messages cannot be delivered (e.g. the target denies access to the service), the initiator never knows about this, since he already ended the conversation. My recomendation is to have the target end the dialog first, after receiving the two messages, and the initiator END CONVERSATION on it's turn as a reaction of receiving the end dialog message from target. Another side effect of this pattern of messages (when target never sends any message back) is that the target endpoint doesn't get deleted, and that is the problem you're seeing. This is a known issue and will be addressed in a next release.

If you're asking about the reason for the 'An error occurred while receiving data: '64(error not found)', this is simply because the connection between the two machines is timing out due to innactivity. One side will time out first (the enterprise in this case) and close the connection, the other side simply reports the fact that the connection was closed by the peer.

HTH,
~ Remus