BPEL process on error retrying self after every 5 minutes (no fault policy)

Question

Hi All
We are facing very strange issue in one of our BPEL processes (we are using Oracle SOA Suite 11.1.1.3 on linux with 2 soa servers in cluster). The composite consists of a web service binding followed by a mediator which routes to three different BPEL processes (operational routing). This is all one way.
Inside the BPEL we are throwing a bpel rollback fault in the catch all to roll back any pending database changes. As soon as the error comes, In the composite diagram i can see multiple BPEL instances under the mediator and they keep on coming after every 5 minutes (basically i see two BPEL instances with a gap of one second and this pair keeps on repeating every 5 minutes)
Also we have a send email in the catch all and the email we get has a different composite id in every mail. However in the em i can see only one composite id (the first one when the error came) with multiple BPEL diagrams in it.
We are not sure what is causing the error?
1) Is it the one way mediator routing to 3 BPEL processes as per the operation invoked?
2) Is there something wrong with clustering?
3) Is there a problem with version 11.1.1.3?
Please help as we are struck and not able to move to INT.
Thanks and Regards
Amit Sachdev

Answer 1

Hi Anuj
Thanks for the reply again. Using the below property seems to work, but it is making the call sync (i get an empty fault in SOAP-UI) and my java client also waits for the response.
<component name="abcd">
<implementation.bpel src="abcd.bpel"/>
<property name="bpel.config.oneWayDeliveryPolicy" many="false" type="xs:string">sync</property>
<property name="bpel.config.transaction" many="false" type="xs:string">required</property>
</component>
As i told that this is a long running process (could easily take 3-4 hours), i cant afford to wait for this (should be fire and forget).
1) Is there a way, i can make it fire and forget and also stop the recovery using the properties you just told?
2) If there is no way out, i can remove the throw rollback fault in catch all, but i want the composite instance to be shown as red (i will take care of the rollbacks in db in code). I tried using the terminate activity but that doesn't make it red (like it used to be in 10g). Is there a way to mark the instance as red without throwing an uncaught fault?
Please let me know. Thanks
Also regarding the patch, if we apply it, will it still recover the BPEL instance? I mean the BPEL is throwing a rollback fault and is being exactly retried 50 times with a gap of 5 minutes (default settings in em console, not touched at all). Every time it runs it would throw the same fault again (as it reads and deletes a control file, so in successive runs that control file will not be there) leading to rollback error being thrown from catch all during automatic retry.