Question

We have developed an application which work on two persistence units. The application retrieves entities from one PU1 and process them in a loop.

In this loop it updates records of PU2 and at the end updates PU1 entities and marks them as processed.

It’s working fine but the problem is I have to begin/commit PU2 for each traverse of the PU1 entitiy due to which we are facing a performance issue.

try{
em1.getTransaction.begin();
Query query = em1.createNamedQuery("Request.SELECT_ALL");
query.setMaxResults(1000);
List<LoaneeRefill> requests = query.getResultList();

for (Request vo : requests) {

 em2.getTransaction.begin();

 Response r =new Resopnse();
 r.setNumber(vo.getNumber);

 em2.persist(r);

 em2.getTransaction.commit();

 vo.setStatus("Y");
 em1.merge(vo);
 }


 em1.getTransaction.commit();
 }catch(Exception e){}

The above code is causing performance issues/taking lots of time on a larger number of requests.

I have tried the following code but got an error:

 try{
 em1.getTransaction.begin();
 em2.getTransaction.begin();

 Query query = em1.createNamedQuery("Request.SELECT_ALL");
 query.setMaxResults(1000);
 List<LoaneeRefill> requests = query.getResultList();

 for (Request vo : requests) {

 Response r =new Resopnse();
 r.setNumber(vo.getNumber);

 em2.persist(r);

 vo.setStatus("Y");
 em1.merge(vo);
 }


 em2.getTransaction.commit();
 em1.getTransaction.commit();
 }catch(Exception e){}

Error:

javax.persistence.PersistenceException: Exception [EclipseLink-7197] (Eclipse Persistence Servi
ces - 2.1.1.v20100817-r8050): org.eclipse.persistence.exceptions.ValidationException
Exception Description: Null or zero primary key encountered in unit of work clone

EDIT:

Any suggestion if I get a list of entities from PU1 and detach them. Then process those requests and update the list with one update or any other way.

Answer 1

It depends on how you are generating the IDs in the database. If they are generated through auto increment on the database, then the information below is likely your only option to get better performance.

If you have another ID system, or can change the way your IDs are generated, take a look here for more info on how ID’s can be generated using tables or sequences, and the benefits and pitfalls of both systems.

---

At the end of your for loop, try adding…

em2.flush();

The reason is JPA does not actually commit the change to the database unit the commit line. Because of this, the ID is not set yet. (I think it adds a 0 there for a placeholder) and when you try to persist a second item without committing the first, it ends up with a second record with the same placeholder id.

The flush will actually execute the SQL to actually add the record to the database and get the real ID created, but leaves the transaction open, so you should see performance gains.

It seems like the persistence API should be smart enough to handle multiple new records at once, but that hasn’t been my experience, there’s likely a good reason, but I don’t know what it is.

---