Example: Single Fault Recovery for Oracle Mediator

8-38 Oracle Fusion Middleware Administrators Guide for Oracle SOA Suite and Oracle BPM Suite Faults identified as recoverable can be recovered. 3. Select faults for recovery. As with fault recovery at the SOA Infrastructure level and BPEL process and Oracle Mediator service component levels, you can perform single fault recovery, bulk fault recovery, and recovery of all faults. See Step 3 of Section 8.5, Recovering from SOA Composite Application Faults at the SOA Infrastructure Level for instructions on selecting faults to perform these types of recovery.

4. Select an action from the Recovery Action list.

Note: You cannot search for human workflow error messages by entering details in the Error Message Contains field because these faults do not persist in the dehydration store. Action Description Action is Available for... Retry Retries the instance directly. An example of a scenario in which to use this recovery action is when the fault occurred because the service provider was not reachable due to a network error. The network error is now resolved. BPEL process and Oracle Mediator Abort Terminates the entire instance. BPEL process and Oracle Mediator Replay Replays the entire scope again in which the fault occurred. BPEL process Managing SOA Composite Applications 8-39

5. If you want to delete rejected messages, click Delete Rejected Messages.

This displays the Delete: Rejected Messages dialog for specifying criteria for deleting rejected messages of the current composite directly from the database.

6. If you want to perform a bulk recovery of messages, click Recover with Options.

This displays the Recover with Options dialog for specifying criteria for recovering BPEL and Oracle Mediator messages of the current composite directly from the database. Human workflow messages can be recovered manually from Oracle BPM Worklist. Business event and business rule messages cannot be recovered. Rethrow Rethrows the current fault. BPEL fault handlers catch branches are used to handle the fault. By default, all exceptions are caught by the fault management framework unless an explicit rethrow fault policy is provided. BPEL process Continue Ignores the fault and continues processing marks the faulted activity as a success. BPEL process Note: In most cases, fault policy actions are automatically executed. The only exception is if you defined a fault policy that uses the action ora-human-intervention. This action creates a recoverable fault that can be recovered from Oracle Enterprise Manager Fusion Middleware Control. Action Description Action is Available for... 8-40 Oracle Fusion Middleware Administrators Guide for Oracle SOA Suite and Oracle BPM Suite

7. Specify criteria. Retry and Abort are the only recovery actions permitted.

8. Click Recover. Depending upon the number of messages, recovery can take some

time. 9. Perform the following additional monitoring tasks from within the faults table:

a. From the View list, select Columns Fault ID to display the fault IDs for each

error message. The fault ID is automatically generated and uniquely identifies a fault. The fault ID is also displayed when you click an error message.

b. In the Fault Location column, click a specific location to access the faults page

for the location of the fault. The location can be a service, component, or reference.

c. In the Component Instance ID column, click a specific service component ID

to access task details about the instance for example, the current state of a task. Note that rejected messages do not have a component instance ID.

d. In the Logs column, click a specific log to access the Log Messages page with

filtered messages specific to that instance. For more information, see the following sections: Note: For bulk fault recovery at the SOA composite application level, a check of the state of the composite is performed. If the state of the composite is set to off, a message is displayed warning you that a recovery cannot be performed. You are not notified when a fault has been skipped during recovery for any reason for example, an unsupported service engine, an unrecoverable fault, and so on.