If a valid XML schema duration or datetime value is returned from the bpelx:for or bpelx:until attribute, a bpelx:timeout fault is thrown from the timed-out activity. This fault can be caught by any catch or catchAll block and handled like a regular BPEL fault. The message of the fault is the name of the activity. In addition, an event is logged to the instance audit trail indicating that the activity has timed out because the expected callback message failed to be received before the timeout duration.

If the activity receives a callback from the partner before the timeout period, no fault is thrown. If a callback is received while the activity is being timed out, the callback message is not delivered to the activity and is marked as canceled in the delivery message table. If a timeout action is attempted at the same time that a callback message is handled, the timeout action is ignored. As of 11g Release 1, instances are locked optimistically (as opposed to pessimistic locking in Release 10g). Therefore, the second action in line is still performed.

The bpelx:timeout fault can be thrown from a BPEL component if the component WSDL declares the fault on the operation. If the fault is not declared on the operation, the fault is converted into a FabricInvocationException runtime fault. This fault can be caught by any caller components (including BPEL components), but the fault type is no longer bpelx:timeout. (However, the fault message string still indicates that the fault was originally a timeout fault.)