Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [eclipselink-users] Workaround? for: EntityManagerFactoryImpl throwing null pointer exception on connection
Yes I tried that, but the close() throws an exception (the other bug). So this variant does not close() the emf. It removes the session 
after closing the EntityManager.
Gordon

Tom Ware wrote:

Where are you removing the session? Is it immediately after the "close"?

-Tom

Gordon Ferguson wrote:

Tom, to follow up on this:
I found the workaround you posted with the bug report and
tried it but still get a socket reset exception when I try
to work with the connection.
Gordon

Gordon Ferguson wrote:

Hi Tom,
I will follow those bugs, thanks.
And I'd be glad for some help with a workaround (if
it's not too complicated).
Thanks,
Gordon

Tom Ware wrote:

Hi Gordon,
I have filed the following bugs. I am going to recommend to the team that we address at both of them for our 1.1 release: 

https://bugs.eclipse.org/bugs/show_bug.cgi?id=256284
https://bugs.eclipse.org/bugs/show_bug.cgi?id=256296
There is a workaround available for the EMF version. I involves using EclipseLink specific API to ensure a copy of the underlying session is not stored after your EMF closes. If you want, I can help you with that workaround. 

-Tom


Gordon Ferguson wrote:

Here you go. I have not reviewed these, hope they help.
Gordon

Tom Ware wrote:

Hi Gordon,
You you increase your logging level to FINEST and resend the output of the failures.
You can user the persistence unit property eclipselink.logging.level=FINEST to set your logging level: 

<property name="eclipselink.logging.level" value="FINEST"/>

-Tom

Gordon Ferguson wrote:

Hi Tom,
Short answer seems to be: the difference is how we
force the diconnection.
I have attached some files to document this.

Upshot is
(1) I was able to replicate your successful restart after an orderly stop/start of MySQL running locally. 
(2) The restart does not succeed when network connectivity to a
remote oracle instance is broken and restored.
(3) The EMF variant, closing and recreating the entire factory,
has different, perhaps more informative, symptoms.

Attached are: the java test file, console output from the success,
console output from the failure, and console output from the
ORACLE failure using the EMF variant. The only difference in the java 
code between the two was to swap persistent unit names.
In the mysql case I used the database manager to stop/start.
In the ORACLE case I disconnected/reconnnected the VPN we use.

The EMF failure had problems with closing the EMF -- it apparently
tried to access the database as part of the close() and failed. And
it apparently did not clear out the old EM:
Caused by: Exception [EclipseLink-28009] (Eclipse Persistence Services - 1.0.2 (Build 20081024)): org.eclipse.persistence.exceptions.EntityManagerSetupException Exception Description: Attempted to redeploy a session named file:/D:/projs/costar-1.3/jclass/-Cloaks-dev-prod without closing it. 

Thanks,
Gordon



Tom Ware wrote:

Hi Gordon,
FYI: By default, EclipseLink will create a connection pool for you. The settings you have below affect the connection limits, but do not affect whether or not there is a connection pool.
I cannot recreate the issue you are having. Can you tell me what is different about what I am doing. 

I am running two simple tests on MySQL.  Here is the pseudo code:

// This is the case where the whole EMF is recreated
    public void testEMFCloseAndOpen(){
        Employee emp = null;
// creates an EMF for PU default and then creates an EM from it 
        EntityManager em = createEntityManager();
        em.find(Employee.class, 1);
        try{
            em.getTransaction().begin();
            // shut down DB prior to this line
            emp = new Employee();
            em.persist(emp);
            em.getTransaction().commit();

        } catch (Exception e){
            // get exception here after a number of retries
            e.printStackTrace();
        }
        getEntityManagerFactory().close();
        // restart the db prior to this line
        em = createEntityManager();
        em.getTransaction().begin();
        emp = new Employee();
        em.persist(emp);
        em.getTransaction().commit();
    }

// this is the case where just the EM is recreated
    public void testEMCloseAndOpen(){
        Employee emp = null;
        EntityManager em = createEntityManager();
        em.find(Employee.class, 1);
        try{
            em.getTransaction().begin();
            // shut down DB prior to this line
            emp = new Employee();
            em.persist(emp);
            em.getTransaction().commit();
        } catch (Exception e){
            // get exception here after a number of retries
            e.printStackTrace();
        }
        em.close();
        // restart the db prior to this line
        em = createEntityManager();
        em.find(Employee.class, 1);
        em.getTransaction().begin();
        emp = new Employee();
        em.persist(emp);
        em.getTransaction().commit();
    }

-Tom

Gordon Ferguson wrote:



Tom Ware wrote:

Hi Gordon,
It sounds like you are using the default EclipseLink connection pool. 


No doubt. I did try with
snip ...
emfProperties.put(JDBC_READ_CONNECTIONS_MIN, "1");
emfProperties.put(JDBC_WRITE_CONNECTIONS_MIN, "1");
// Ensure that no server-platform is configured
emfProperties.put(TARGET_SERVER, TargetServer.None);
etc
But the problem remained.
Is there any particular reason you have chosen to close your EntityManagerFactory? What happens if you simply acquire a new entityManager from the same factory? Do you have problems with the connection?
I did try some variations. Closing and recreating everything seemed the most likely way to get a fresh start. Shot in the dark in other words.
I have just (re)tried: close the EntityManager and get a new one from 
the existing factory; the problem appears unchanged.

One, perhaps interesting, note about the stack trace: EL logs
a successful login before throwing the null pointer exception.
I don't see the login success when the underlying network
connection is unavailable.

Thanks again,
Gordon


-Tom

Gordon Ferguson wrote:



Tom Ware wrote:

Hi Gordon,

  Are you running in JavaSE or in some kind of container?

JavaSE


  Are you using a connection pool?

not as far as I know. Should I be?
Is it external to EclipseLink or EclipseLink's own connection pool? 
n/a?


  How does your test obtain the connection and reestablish it?
>> emf = Persistence.createEntityManagerFactory(persistenceUnitLabel); 
 >>   entityManager = emf.createEntityManager();
I am not surprised that the scenario you are describing is causing issues. EclipseLink has some strategies you can use to reestablish connections, but doing it somewhere EclipseLink is completely unaware of could cause an issue.
So, what is the most simple solution? The actual application runs in Tomcat; database is sometimes Oracle sometimes MySQL. (The JNDI stuff seems to multiply code complexity, do I really have to go there? ) 

Many thanks,
Gordon



-Tom

Gordon Ferguson wrote:

Hi Tom,
It has been a few weeks but I'm back to this question.

The basic scenario I'm having trouble with is recovery after a
SQLException: Io exception: The Network Adapter could not establish the connection ...
My test case forces that error and then restores the underlying 
connection. I then want to get EclipseLink going again.
Is there an api to get a (new?) working EntityMangager at this point?
What I'm trying is invoke close() both the EntityManager and the EntityManagerFactory. (Nothing thrown by this.) Then I repeat this code:
protected void privateInit(String persistenceUnitLabel, ...) { emf = Persistence.createEntityManagerFactory(persistenceUnitLabel); entityManager = emf.createEntityManager(); 
  ...
}

and the first line creates the stack trace below.

Thanks for any suggestions,
Gordon

Tom Ware wrote:

Hi Gordon,
If EclipseLink is following the expected code path, the structConverters should be initialized inside the "if (state == STATE_PREDEPLOYED) {" code block below where the structConverters variable is being defined on the following line: 

structConverters = processor.getStructConverters();
If you are not getting into that if statement and still getting to "addStructConverters(session, structConverters)" something strange is occurring with your deployment. It is hard to say what the problem is without more information about the symptoms of the problem you are seeing. 

-Tom

Gordon Ferguson wrote:

Hello,
I'm using the SE approach and testing my 'recovery after error' logic. I see the exception listed below. I looked at the code and found 
EntityManagerSetupImpl.deploy() method initializes
structConverters to null in l.189
Then, in l. 319 passes that null to addStructConverters(...) which tries to add to the (null) List. 
 List<StructConverter> structConverters = null;
SO, I'm wondering if this is just a coding typo or if I need to dig deeper 
into why I'm in this state?

Any suggestions would be appreciated.

Thanks,
Gordon







Listing:
[EL Info]: 2008.10.30 11:53:03.781--ServerSession(653967)--EclipseLink, version: Eclipse Persistence Services - 1.0.2 (Build 20081024) Exception in thread "main" [EL Info]: 2008.10.30 11:53:10.328--ServerSession(653967)--file:/D:/projs/costar-1.3/jclass/-True-DevDev login successful costar.dbaccess.DbAccessException: Could not access database. at costar.dbaccess.jpa.CoStarJpaAccess.privateInit(CoStarJpaAccess.java:82) at costar.dbaccess.jpa.CoStarJpaAccess.<init>(CoStarJpaAccess.java:53) at costar.dbaccess.jpa.test.JpaConnectionCheckerChecker.main(JpaConnectionCheckerChecker.java:104) 
Caused by: java.lang.NullPointerException
at org.eclipse.persistence.internal.jpa.EntityManagerSetupImpl.addStructConverters(EntityManagerSetupImpl.java:319) at org.eclipse.persistence.internal.jpa.EntityManagerSetupImpl.deploy(EntityManagerSetupImpl.java:249) at org.eclipse.persistence.internal.jpa.EntityManagerFactoryImpl.getServerSession(EntityManagerFactoryImpl.java:69) at org.eclipse.persistence.internal.jpa.EntityManagerFactoryImpl.createEntityManagerImpl(EntityManagerFactoryImpl.java:118) at org.eclipse.persistence.internal.jpa.EntityManagerFactoryImpl.createEntityManagerImpl(EntityManagerFactoryImpl.java:112) at org.eclipse.persistence.internal.jpa.EntityManagerFactoryImpl.createEntityManager(EntityManagerFactoryImpl.java:100) at costar.dbaccess.jpa.CostarJpaFacade.<init>(CostarJpaFacade.java:47) at costar.dbaccess.jpa.CoStarJpaAccess.privateInit(CoStarJpaAccess.java:80) 
    ... 2 more

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users


_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users


_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users


_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users
------------------------------------------------------------------------ 

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users
------------------------------------------------------------------------ 

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users


_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users



------------------------------------------------------------------------

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-users

Back to the top