[rules-users] JBPM fails to resume a rehydrated active process instance

Fri Jun 8 02:37:15 EDT 2012

Finally found the problem.

JBPM's WSHumanTaskHandler connects to JBPM HumanTask server only when
executing a workitem. If a connection is already established it doesn't
connect again so it is durable. At connection time, it also registers
taskEventHandlers, one of them for receiving TaskCompleted events.

In a JPA persisted environment where a rehydrated knowledge session which
has pending workitems with tasks living in the human task server has been
loaded, if a task get completed at the human task server, there's nobody
listening for events because JBPM doesn't connect to the server until a new
HumanTask workitem is executed, so the process doesn't continue it's
execution. This behaviour will continue to happen until a new workitem gets
executed, then the connection will be stablished, callback handlers
registered, and from that point on workitems, past and future, get
completed and the process execution resumed.

Not without problems, when querying the knowledge session model a new
exception arises:

javax.persistence.TransactionRequiredException: No active JTA transaction
on joinTransaction call
   at
org.hibernate.ejb.AbstractEntityManagerImpl.joinTransaction(AbstractEntityManagerImpl.java:1183)
   at
org.hibernate.ejb.AbstractEntityManagerImpl.joinTransaction(AbstractEntityManagerImpl.java:1143)
   at
org.drools.persistence.jpa.JpaPersistenceContextManager.beginCommandScopedEntityManager(JpaPersistenceContextManager.java:83)

But that's another thing. :(

Created a Jira issue for this:  https://issues.jboss.org/browse/JBPM-3673

Alberto R. Galdo
argaldo at gmail.com

On Tue, Jun 5, 2012 at 8:44 AM, Alberto R. Galdo <argaldo at gmail.com> wrote:

>
> Sorry, I didn't mention ... We're using the latest stable version of both
> Drools, JBPM and Knowledge API ... Drools 5.4.0.Final, JBPM 5.3.0.Final
> mostly because they're dependent on each other ... :(
>
> Any hints on the subject?
>
> Alberto R. Galdo
> argaldo at gmail.com
>
>
>
>
> On Mon, Jun 4, 2012 at 6:02 PM, Edson Tirelli <ed.tirelli at gmail.com>wrote:
>
>>
>>    Alberto,
>>
>>    Which versions of Drools and jBPM are you using?
>>
>>    Edson
>>
>> On Mon, Jun 4, 2012 at 3:55 AM, Alberto R. Galdo <argaldo at gmail.com>wrote:
>>
>>> Hi guys,
>>>
>>>    We are using JBPM 5 for long running processes in one of our
>>> products, both, the knowledge session and WorkItemInfo &
>>> ProcessInstanceInfo are persisted in a PostgreSQL database. Our processes
>>> involve a set of automated tasks along with UserTasks handled inside JBPM's
>>> HumanTask provided server  ( the one that uses mina & JPA ).
>>>
>>>    For maintenance purposes, our application needs to be restarted once
>>> in a while so we need our processes to keep running after a sucesful
>>> rehydratation of the knowledge & JBPM contexts . We've tested a persistent
>>> approach like this:
>>>
>>>      - We started our application and started several instances of a
>>> process that involves at least one human task node.
>>>      - We've completed several human tasks in the current session and
>>> things went ok, workitems got completed & processinstances desappeared as
>>> processes came to and end.
>>>      - We left several human task in state"Ready" for our users to
>>> consume afterwards.
>>>      - We restarted the application and let JBPM rehydrate it's context
>>> via   JPAKnowledgeService.loadStatefulKnowledgeSession()
>>>
>>>    Then we acquired and completed a "Ready" HumanTask, it got completed
>>> in JBPM HumanTask server's database backend but the process attached never
>>> got neither updater, nor its subsequent nodes were executed. This is weird.
>>>
>>>     We first tought it was a rehydratation problem and that the process
>>> definition was not loaded properly, but what was our suprise when we've
>>> been able to execute several new process instances wich we could complete
>>> properly. So, seems that we are able to complete "current" ( as in the
>>> current session, not the rehydrated ones ) process instances, but not able
>>> to complete process instances that are not started in the "current" session
>>> ( the rehydrated ones ).
>>>
>>>    What's wrong here. Is it not possible to resume a process in a
>>> rehydrated session when human tasks & JBPM's HumanTask server are involved?
>>>
>>>
>>> Greets,
>>>
>>> Alberto R. Galdo
>>> argaldo at gmail.com
>>>
>>>
>>>
>>> _______________________________________________
>>> rules-users mailing list
>>> rules-users at lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/rules-users
>>>
>>>
>>
>>
>> --
>>   Edson Tirelli
>>   JBoss Drools Core Development
>>   JBoss by Red Hat @ www.jboss.com
>>
>> _______________________________________________
>> rules-users mailing list
>> rules-users at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/rules-users
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-users/attachments/20120608/36d91a1d/attachment-0001.html 