Transactional behaviour in third party -> XI idoc scenario

Hi,

a third party system is sending an idoc via TRFC-call with the function module "IDOC_INBOUND_ASYNCHRONOUS" to the XI-system.

The scenario is working fine, but one day the XI table space was full.

The third party systems gets the following error message:

<b>RFC_ERROR_SYSTEM_FAILURE Problems found during the EXPORT to specified key in table.</b>

For the sending third party system that is an error and the internal status is set to mode "rollback". That means the third party systems is going to send the idocs again after a specified period of time.

If i look in transaction SXMB_MONI i can see that the inbound queues were stopped (XI tried to open more than one queue, but all queues were stopped).

After getting more table space available for XI all (stopped) inbound queues were activated manually, and <b> now all messages waiting in XI were processed!</b>. To me that means XI was able to receive all messages, even the table space was full???!!!!

But this behaviour is wrong in a transactional view of the whole process. The sending system is going to send the idocs again and so the idoc's were processed twice in the XI system and respective in the receiver system.

How to handle that?