Hello Developers, Strong suggested I fwd this to the mailing-list.With reference to mail below could any of you clarify/explain if this is expected behaviour of NaturalIdLoadAccess with 2lcache as I find some discrepancy. Thanks and Regards, Madhumita Sadhukhan JBoss EAP QE Team Red Hat Brno ----- Forwarded Message ----- From: "Madhumita Sadhukhan"<msadhukh(a)redhat.com> To: "Strong Liu"<stliu(a)redhat.com> Sent: Friday, April 27, 2012 12:19:18 AM Subject: NaturalIdLoadAccess behaviour is this expected? Hi Strong, I notice a strange behaviour while loading natural id with secondlevel cache enabled. Not sure if this is expected behaviour but I notice some discrepancies which I would like to clarify. I am not yet uploading this test to my AS7 branch on github as AS7 is still stuck on hibernate4.1.2 and this test requires 4.1.3 from EAP 6 tested currently to work. Please paste the attached unzipped folder in AS testsuite folder structure in location: /jboss-as/testsuite/integration/basic/src/test/java/org/jboss/as/test/integration/hibernate/ Please run with hibernate 4.1.3 jars replaced in modules/org/hibernate..... within build of AS7 In my test I have tried to load an entity with NaturalIdLoadAccess while 2lcache is enabled using two natural ids(firstname and voterid) in several steps. 1) create person(with natural ids firstname and voterid) 2) load using natural ids first time 3) modify/update natural ids in database(in order to not touch 2lcache) 4) load using loader from 2Lcache using old natural id values in step 2---//this works correctly as expected and is able to load values from cache though it has been modified in DB. 5) then load using loader from (2) but using updated values of firstname and voterid(naturalids) used in step 3 //this is where it breaks the Person object returned from this load still shows older value of firstname as in step 2 but wondering how the loader works as I have passed the new values in using(....) also the next step fails so wondering where the flaw is! 6) try to use same loader and load using older values of natural id this is to recheck if older value is still persisting in cache as it was returned in step5 //this returns null pointer showing it does not exist in cache and loader is unable to load it! 7) If i try again with the new actual values the loader is able to load the person entity again but with old values!!! So the problem/confusion is due to the discrepancy between step 5 and step 6 (with the older value of entity loaded in step5 indicating it is still cached but then with the failure to load it in step 6 throwing nullpointer which indicates its not) Also I am surprised at how using() function behaves on the loader i.e it returns the entity with older values while loading "using()" the updated values!(as confirmed in step 7) Please note that the testcase will pass as I replaced the asserts with S.O.Ps hence you should check for the values in server log excerpt(attaching for ease) Could you take a quick look please and explain if this is expected behaviour? Thanks and Regards, Madhumita Sadhukhan JBoss EAP QE Team Red Hat Brno _______________________________________________ hibernate-dev mailing list hibernate-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/hibernate-dev

Steve, I did attach the piece of code and log and its showing in my sent items.I am surprised why the attachment did not make it(as the size is few kbs). I will try again and I hope it makes it this time(attached files) . And by loader I meant an instance of NaturalIdLoadAccess(which will be clear if you can take a look at the code). Thanks and Regards, Madhumita Sadhukhan JBoss EAP QE Team Red Hat Brno ----- Original Message ----- From: "Steve Ebersole"<steve(a)hibernate.org> To: "Madhumita Sadhukhan"<msadhukh(a)redhat.com> Cc: "hibernate-dev(a)lists.jboss.org hibernate-dev"<hibernate-dev(a)lists.jboss.org> Sent: Friday, April 27, 2012 4:47:33 PM Subject: Re: [hibernate-dev] Fwd: NaturalIdLoadAccess behaviour on 2Lcache is this expected? I have no idea what "load by loader" means. "Loader" to Hibernate developers means something much different than I am assuming you mean... Also, I see no attachment... First and foremost, I'd argue that your use case here is just plain broken. Just like with entity data, second-level caching data that you fully expect to change "under the covers" is just asking for problems. That being said, we should certainly try to handle the situation here *it if it makes sense*. But thats hard to ascertain without the logs or the mapping or something. I assume the entity itself is also cached? Thats the only way I can make sense out of what I think you are doing in (5) On Fri 27 Apr 2012 09:14:31 AM CDT, Madhumita Sadhukhan wrote: > Hello Developers, > > Strong suggested I fwd this to the mailing-list.With reference to mail below could any of you clarify/explain if this is expected behaviour of NaturalIdLoadAccess with 2lcache as I find some discrepancy. > > Thanks and Regards, > Madhumita Sadhukhan > JBoss EAP QE Team > Red Hat Brno > > > > > ----- Forwarded Message ----- > From: "Madhumita Sadhukhan"<msadhukh(a)redhat.com> > To: "Strong Liu"<stliu(a)redhat.com> > Sent: Friday, April 27, 2012 12:19:18 AM > Subject: NaturalIdLoadAccess behaviour is this expected? > > Hi Strong, > > I notice a strange behaviour while loading natural id with secondlevel cache enabled. > Not sure if this is expected behaviour but I notice some discrepancies which I would like to clarify. > I am not yet uploading this test to my AS7 branch on github as AS7 is still stuck on hibernate4.1.2 and this test requires 4.1.3 from EAP 6 tested currently to work. > Please paste the attached unzipped folder in AS testsuite folder structure in location: > > /jboss-as/testsuite/integration/basic/src/test/java/org/jboss/as/test/integration/hibernate/ > > Please run with hibernate 4.1.3 jars replaced in modules/org/hibernate..... within build of AS7 > > In my test I have tried to load an entity with NaturalIdLoadAccess while 2lcache is enabled using two natural ids(firstname and voterid) in several steps. > > 1) create person(with natural ids firstname and voterid) > 2) load using natural ids first time > 3) modify/update natural ids in database(in order to not touch 2lcache) > 4) load using loader from 2Lcache using old natural id values in step 2---//this works correctly as expected and is able to load values from cache though it has been modified in DB. > 5) then load using loader from (2) but using updated values of firstname and voterid(naturalids) used in step 3 > //this is where it breaks the Person object returned from this load still shows older value of firstname as in step 2 but wondering how the loader works as I have passed the new values in using(....) > also the next step fails so wondering where the flaw is! > > 6) try to use same loader and load using older values of natural id this is to recheck if older value is still persisting in cache as it was returned in step5 > //this returns null pointer showing it does not exist in cache and loader is unable to load it! > > 7) If i try again with the new actual values the loader is able to load the person entity again but with old values!!! > > > So the problem/confusion is due to the discrepancy between step 5 and step 6 (with the older value of entity loaded in step5 indicating it is still cached but then with the failure to load it in step 6 throwing nullpointer which indicates its not) > Also I am surprised at how using() function behaves on the loader i.e it returns the entity with older values while loading "using()" the updated values!(as confirmed in step 7) > > Please note that the testcase will pass as I replaced the asserts with S.O.Ps hence you should check for the values in server log excerpt(attaching for ease) > > > Could you take a quick look please and explain if this is expected behaviour? > > > Thanks and Regards, > Madhumita Sadhukhan > JBoss EAP QE Team > Red Hat Brno > > > > > > > _______________________________________________ > hibernate-dev mailing list > hibernate-dev(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/hibernate-dev -- steve(a)hibernate.org http://hibernate.org

Thank you Madhumita , finally I was able to reproduce all steps. Whilst point 6 represents definitely a bug (I opened https://hibernate.onjira.com/browse/HHH-7287) I'm not so sure about point 5: the NaturalIdLoadAccess's internal query does 'see' the new values, when the used isolation level is lesser than SERIALIZABLE and therefore return a result. The question here is something to discuss , in my opinion,is. Is whether: - NaturalIdLoadAccess should act like the persistent context's repeatable-read behavior, and thus return null in case that the search-natural-id-values differ from the ones in the according entity state. or if - NaturalIdLoadAccess should act like a plain query api and thus return the entity even if the search-natural-id-values differ from the ones in the current entity state. best regards Guenther -----Original Message----- From: Madhumita Sadhukhan [mailto:msadhukh@redhat.com] Sent: Monday, April 30, 2012 11:32 AM To: Demetz, Guenther; Steve Ebersole Cc: hibernate-dev(a)lists.jboss.org hibernate-dev Subject: Re: [hibernate-dev] Fwd: NaturalIdLoadAccess behaviour on 2Lcache is this expected? Hello Steve and Guenther, Thanks for sparing some time to look into this. The entitymapping information for Person is in file HibernateNativeAPINaturalId2LCachingTestCase.java in string testmapping. For convenience I have uploaded the test to my AS7 github as follows: https://github.com/madhumita12/jboss-as/blob/master/testsuite/integration... Files you need to look at: HibernateNativeAPINaturalId2LCachingTestCase.java SFSB2LcacheNaturalId.java Person.java This testcase will fail in this branch as it uses hibernate 4.1.2 and in order to run it successfully you could just replace with 4.1.3 hibernate jars after building this branch. I have tried to put adequate comments to express my location of confusion. Steve, I am following the normal pattern followed by most jpa/hibernate integration testcases in AS7 where the methods/actions involving the Transaction or Session is handled in the (SF/SL)Session Bean which in turn is invoked from the main testcase that prepares the archive and deploys the ear ,invokes the tests with conclusions in assertions. Guenther, I agree with your reasoning for steps 5 and step 7 as I am using Container Managed Transaction where isolation level (if it is not defined) picks up the default for the underlying Database which is inbuilt H2 for AS7 in this case and default isolation used for H2 is read-committed. However as I have already explained in my previous mail I am rather confused/concerned with the discrepancy in behaviour between step 5 and step 6 as step 6 returns me null when I use the same NaturalIdLoadAccess object to load the entity using previous values which looks inconsistent! Thanks and Regards, Madhumita Sadhukhan JBoss EAP QE Team Red Hat Brno ----- Original Message ----- From: "Guenther Demetz"<Guenther.Demetz(a)wuerth-phoenix.com> To: "Madhumita Sadhukhan"<msadhukh(a)redhat.com>, "hibernate-dev(a)lists.jboss.org hibernate-dev"<hibernate-dev(a)lists.jboss.org> Sent: Monday, April 30, 2012 9:16:44 AM Subject: RE: [hibernate-dev] Fwd: NaturalIdLoadAccess behaviour on 2Lcache is this expected? Hi Madhumita Sadhukhan, although I did not receive your attachments, please allow me to make some considerations to point 5 and successive: point 5: When you load using updated natural-ids values, it should depend on the isolation level you use, whether you retrieve a result or not. With levels REPEATABLE_READ of SERIALIZEABLE there should be no result, since the update-transaction begun after you already read concerning record in your main transaction. With READ_COMMITED or lower isolation you have a result, but this will not update the values in your persistent context as it simulates repeatable-read for entities. This is a well-documented behavior, explained in book Java Persistence with Hibernate, chapter 9.1.2 The persistent context, here a snippet from: ... Hibernate reads the result set of the query and marshals entity objects that are then returned to the application. During this process, Hibernate interacts with the current persistence context. It tries to resolve every entity instance in this cache (by identifier); only if the instance can’t be found in the current persistence context does Hibernate read the rest of the data from the result set. It tries to resolve every entity instance in this cache (by identifier); only if the instance can’t be found in the current persistence context does Hibernate read the rest of the data from the result set. This may explain the behavior in point 5 and 7, does it? (Indeed I was not able to reproduce point 6, I'm not getting a null as result.) regards G.D. -----Original Message----- From: hibernate-dev-bounces(a)lists.jboss.org [mailto:hibernate-dev-bounces@lists.jboss.org] On Behalf Of Madhumita Sadhukhan Sent: Friday, April 27, 2012 4:15 PM To: hibernate-dev(a)lists.jboss.org hibernate-dev Subject: [hibernate-dev] Fwd: NaturalIdLoadAccess behaviour on 2Lcache is this expected? Hello Developers, Strong suggested I fwd this to the mailing-list.With reference to mail below could any of you clarify/explain if this is expected behaviour of NaturalIdLoadAccess with 2lcache as I find some discrepancy. Thanks and Regards, Madhumita Sadhukhan JBoss EAP QE Team Red Hat Brno ----- Forwarded Message ----- From: "Madhumita Sadhukhan"<msadhukh(a)redhat.com> To: "Strong Liu"<stliu(a)redhat.com> Sent: Friday, April 27, 2012 12:19:18 AM Subject: NaturalIdLoadAccess behaviour is this expected? Hi Strong, I notice a strange behaviour while loading natural id with secondlevel cache enabled. Not sure if this is expected behaviour but I notice some discrepancies which I would like to clarify. I am not yet uploading this test to my AS7 branch on github as AS7 is still stuck on hibernate4.1.2 and this test requires 4.1.3 from EAP 6 tested currently to work. Please paste the attached unzipped folder in AS testsuite folder structure in location: /jboss-as/testsuite/integration/basic/src/test/java/org/jboss/as/test/integration/hibernate/ Please run with hibernate 4.1.3 jars replaced in modules/org/hibernate..... within build of AS7 In my test I have tried to load an entity with NaturalIdLoadAccess while 2lcache is enabled using two natural ids(firstname and voterid) in several steps. 1) create person(with natural ids firstname and voterid) 2) load using natural ids first time 3) modify/update natural ids in database(in order to not touch 2lcache) 4) load using loader from 2Lcache using old natural id values in step 2---//this works correctly as expected and is able to load values from cache though it has been modified in DB. 5) then load using loader from (2) but using updated values of firstname and voterid(naturalids) used in step 3 //this is where it breaks the Person object returned from this load still shows older value of firstname as in step 2 but wondering how the loader works as I have passed the new values in using(....) also the next step fails so wondering where the flaw is! 6) try to use same loader and load using older values of natural id this is to recheck if older value is still persisting in cache as it was returned in step5 //this returns null pointer showing it does not exist in cache and loader is unable to load it! 7) If i try again with the new actual values the loader is able to load the person entity again but with old values!!! So the problem/confusion is due to the discrepancy between step 5 and step 6 (with the older value of entity loaded in step5 indicating it is still cached but then with the failure to load it in step 6 throwing nullpointer which indicates its not) Also I am surprised at how using() function behaves on the loader i.e it returns the entity with older values while loading "using()" the updated values!(as confirmed in step 7) Please note that the testcase will pass as I replaced the asserts with S.O.Ps hence you should check for the values in server log excerpt(attaching for ease) Could you take a quick look please and explain if this is expected behaviour? Thanks and Regards, Madhumita Sadhukhan JBoss EAP QE Team Red Hat Brno

Whilst point 6 represents definitely a bug (I opened https://hibernate.onjira.com/browse/HHH-7287)

>Anyway, like I said in one of my earlier replies, Hibernate could >certainly support this use case if thats what we all decide it should. >But I will limit this to natural ids mapped as mutable. The reason >being that there will be alot of overhead in supporting this correctly >and it is really only relevant for mutable natural ids. Developers who >have developed their apps to leverage non-mutable natural ids should not >suffer in this perf hit.

> If you change your non-transactional access to instead be transactional, it works.

>Your in-line comments mention something about SERIALZABLE txn isolation, but >this is against H2 and just using (default) READ COMMITTED.

>Anyway, like I said in one of my earlier replies, Hibernate could >certainly support this use case if thats what we all decide it should. >But I will limit this to natural ids mapped as mutable. The reason >being that there will be alot of overhead in supporting this correctly >and it is really only relevant for mutable natural ids. Developers who >have developed their apps to leverage non-mutable natural ids should not >suffer in this perf hit.