Thanks for response! By terminating them, I mean calling the Instance Management terminate operation to kill them. A lot of our BPEL services are synchronous, read-only operations and we really don't need them around.
I've used the cleanup parameters in deploy.xml, but found them to be horrifically slow. As in, it upped the execution time from 1 second to 5 seconds. We are using the embedded Derby database - it may be time to move to a "real" database back end. Is that a common scenario for production environments? It just doesn't sound right to me - we don't have extremely high volume. Is there some optimization we can do to the ODE database? Add indexes perhaps? -----Original Message----- From: Antoine Toulme [mailto:[email protected]] Sent: Tuesday, January 19, 2010 11:54 AM To: [email protected] Subject: Re: Recommended Cleanup Procedure? Hi, please see my comments below: On Tue, Jan 19, 2010 at 05:51, Craig Riecke <[email protected]> wrote: > OK, so after successfully redeploying a BPEL process (through the API, so > versioning is used), there's a retired process lying around. It might have > some active processes that we need to get rid of. What is the recommended > procedure for getting rid of this retired process? My thought is: > > > * Terminate all the non-errored, non-completed instances. > You mean, kill them ? Or let them finish ? > > * Delete all the instances of this process, which should now > include all the terminated ones. This seems to take a long, long time > through the Web Service API. > You can use the cleanup parameters in your deploy.xml to remove instance data/ > > * Now there should be no instances left. Undeploy the retired > instance. > It's not always right to undeploy the previous version. You might want to reactivate it later. In any case, this would not be automatic I think. > > Does that sound right? Is there a more automatic way to flush retired > processes? (Ideally I'd like to set a cron job that says "flush all retired > processes after 7 days.") >
