Grouper Developers Forum

[Jim Fox <>]

Enabling them works for me.  And that what I've done.  So I'm fine.

There is a note in the properties file that seems to indicate nested 
transactions are sometimes harmful and can be turned off without causing any 
trouble.  I don't really understand the situation well enough to know why I 
would either enable or disable them.  Maybe it's a documentation issue.

Jim

On Dec 15, 2012, at 10:55 AM, Chris Hyzer wrote:

> Jim,
> 
> Thanks for your research in this one, it really helps a lot.  Im trying to 
> figure out what is happening, and basically the post commit hooks run in a 
> transaction trigger, which actually runs in the same HibernateSession.  So 
> if you don't have nested transactions, then it thinks it is the same 
> HibernateSession, but that transaction is closed.  So... I think if you 
> have a post commit hook you cant turn nested transactions off.  I don't 
> think you should turn that off in general.  Anyways, I think we could fix 
> this somehow with some complicated code, but I don't really think it is 
> worth it, just enabled nested transactions...  ok?  :)
> 
> Thanks,
> Chris
> 
> -----Original Message-----
> From: Jim Fox 
> [mailto:]
>  
> Sent: Thursday, December 13, 2012 4:00 PM
> To: Chris Hyzer
> Cc: Grouper Dev
> Subject: RE: [grouper-dev] 2.1.2 hook writer not writing?
> 
> 
> 
> Here's the trouble.
> 
> There is new code in the HibernateSession constructor:
> 
>     //if we arent using nested transactions, then just use parent if there 
> is one...
>     if 
> (!GrouperConfig.getPropertyBoolean("ddlutils.use.nestedTransactions", true) 
> && parentHibernateSession != null) {
>       grouperTransactionType = 
> parentHibernateSession.getGrouperTransactionType();
>       //we dont want new transactions... not sure what happens if none... 
> hmm
>       if (grouperTransactionType.isNewAutonomous()) {
>         if (grouperTransactionType == 
> GrouperTransactionType.READ_WRITE_NEW) {
>           grouperTransactionType = 
> GrouperTransactionType.READ_WRITE_OR_USE_EXISTING;
>         } else if (grouperTransactionType == 
> GrouperTransactionType.READONLY_NEW) {
>           grouperTransactionType = 
> GrouperTransactionType.READONLY_OR_USE_EXISTING;
>         }
>       }
>       LOG.debug("Not using nested transactions, converting transaction type 
> to: " + parentHibernateSession.getGrouperTransactionType());
>     }
> 
> 
> That's causing the sessions obtained by the hook loop to be reused parents,
> instead of new.  The parent's session appears to be already committed.
> 
> If I set "ddlutils.use.nestedTransactions = true", to bypass the new code
> then I get new sessions and the hooks automatically complete.
> 
> Don't know the reason for the new code, but it might need another check
> somewhere to see if the parent should really be reused.
> 
> Jim
> 
> 
> 
> 
> On Thu, 13 Dec 2012, Jim Fox wrote:
> 
>> Date: Thu, 13 Dec 2012 11:54:18 -0800
>> From: Jim Fox 
>> <>
>> To: Chris Hyzer 
>> <>
>> Cc: Grouper Dev 
>> <>
>> Subject: RE: [grouper-dev] 2.1.2 hook writer not writing?
>> 
>> 
>> And some more:
>> 
>> If I use HibernateSession's toString,
>> 
>> With 2.1.2 (and postgres version 9) I see,
>> (right after the transaction is created and just before executing the hook)
>> 
>>  HibernateSession hibernateSession = 
>> hibernateHandlerBean.getHibernateSession();
>> 
>> log of hibernateSession.toString() shows: isNew: false
>> 
>> But with 2.0.0 (postgres 8) I see in the same place
>> 
>>  HibernateSession hibernateSession = 
>> hibernateHandlerBean.getHibernateSession();
>> 
>> log of hibernateSession.toString() shows: isNew: true
>> 
>> 
>> So, for some reason the session created in that hook loop
>> is no longer a new session.
>> 
>> 
>> Jim
>> 
>> 
>> On Wed, 12 Dec 2012, Jim Fox wrote:
>> 
>>> Date: Wed, 12 Dec 2012 13:38:42 -0800
>>> From: Jim Fox 
>>> <>
>>> To: Chris Hyzer 
>>> <>
>>> Cc: Grouper Dev 
>>> <>
>>> Subject: RE: [grouper-dev] 2.1.2 hook writer not writing?
>>> 
>>> 
>>> Also,  in that same code, right after
>>> 
>>> HibernateSession hibernateSession = 
>>> hibernateHandlerBean.getHibernateSession();
>>> 
>>> hibernateSession.isTransactionActive()  is false
>>> 
>>> and after executeHook it is still false.
>>> 
>>> 
>>> I've tried this with my web service and also with gsh.
>>> --------------------------------
>>> root=GrouperSession.startRootSession()
>>> g=GroupFinder.findByName(root, "u:fox:browser6")
>>> mset=g.getMembers(); marr=mset.toArray()
>>> g.deleteMember(marr[0])
>>> --------------------------------
>>> 
>>> Same result either way.
>>> 
>>> 
>>> Jim
>>> 
>>>> 
>>>> 
>>>> I added some logging and a try/catch around the executeHook and commit 
>>>> statements.
>>>> 
>>>> Jim
>>>> 
>>>> 
>>>>> 
>>>>> I will look when I get home (where I had it running), though in the 
>>>>> meantime, when I look at 2.1.2 source (attached), I don't see log lines 
>>>>> on the line numbers listed.  E.g. GrouperHooksUtils 498 and 504...  any 
>>>>> ideas why?  :)
>>>>> 
>>>>> Thanks,
>>>>> Chris
>>>>> 
>>>>> -----Original Message-----
>>>>> From: Jim Fox 
>>>>> [mailto:]
>>>>> Sent: Wednesday, December 12, 2012 3:35 PM
>>>>> To: Chris Hyzer
>>>>> Cc: Grouper Dev
>>>>> Subject: RE: [grouper-dev] 2.1.2 hook writer not writing?
>>>>> 
>>>>> 
>>>>> 
>>>>> Here's what I'm seeing.
>>>>> 
>>>>> 
>>>>> In GrouperHooksUtils.schedulePostCommitHooksIfRegistered
>>>>> 
>>>>> 
>>>>> There is the call to execute the hook
>>>>> 
>>>>>    executeHook(method, hook, finalHooksBean, hooksContext, null, 
>>>>> asychronous);
>>>>> 
>>>>> followed by a call to commit
>>>>> 
>>>>>    hibernateSession.commit(GrouperCommitType.COMMIT_NOW);
>>>>> 
>>>>> [ I put an exception catch around the commit. ]
>>>>> 
>>>>> 
>>>>> The first statement executes fine, debug log shows:
>>>>> 
>>>>>   GrouperHooksUtils.executeHook(498) - START: Hook 
>>>>> UWMembershipHooks.membershipPostCommitAddMember id: RKW6SJT1
>>>>>   HibernateSession.<init>(151) - Not using nested transactions, 
>>>>> converting transaction type to: READ_WRITE_NEW
>>>>>   GrouperHooksUtils.executeHook(504) - END (normal): Hook 
>>>>> UWMembershipHooks.membershipPostCommitAddMember id: RKW6SJT1 (9ms)
>>>>> 
>>>>> and the spy log shows:
>>>>> 
>>>>>   1ms, statement: Method.java.invoke() line 601, 
>>>>> GrouperUtil.java.invokeMethod() line 3896,
>>>>>      GrouperHooksUtils.java.executeHook() line 502, 
>>>>> GrouperHooksUtils.java.access$100() line 56,
>>>>>      GrouperHooksUtils.java.callback() line 437, 
>>>>> GrouperHooksUtils.java.afterCompletion() line 427,
>>>>>      Group.java.internal_addMember() line 1115, Group.java.addMember() 
>>>>> line 942, Group.java.addMember() line 861
>>>>>   insert into uw_activity (time, action, group_id, data, actors) values 
>>>>> (1355343520114, 'pm', 'a1681c3fcba3f54f759e6c9432004381', 'uwpds imf', 
>>>>> 'fox|fox')
>>>>> 
>>>>> 
>>>>> Then the commit statement fails:
>>>>> 
>>>>>   GrouperHooksUtils$1$1$1.callback(444) - commit: 
>>>>> org.hibernate.TransactionException: Transaction not successfully started
>>>>> 
>>>>> 
>>>>> So the commit fails and the next thing spy shows is a rollback.
>>>>> 
>>>>> If I do the commit manually, in the hook, with
>>>>> 
>>>>>  HibernateSession.bySqlStatic().executeSql( "commit");
>>>>> 
>>>>> That works.
>>>>> 
>>>>> Jim
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>> 
>>> 
>>