topic Re: Issue with Job executioner removing the job record on server restart in Alfresco Archive

Issue with Job executioner removing the job record on server restart

satramalin — Wed, 11 Feb 2015 14:16:51 GMT

Hi,My process has a polling service which has a system task and an user task. The user task has a timer for 3 minutes which routes back to the system task which reads mail. The server hosting the web application had a restart when the job executioner had picked up the job from the jobs table (Now th

Re: Issue with Job executioner removing the job record on server restart

jbarrez — Wed, 18 Feb 2015 15:33:13 GMT

Im not sure I follow. If a job was picked up, and the server went down, the transaction is rolled back in the database, and it is set to the previous state.

So I'm not understanding it correctly, or is it more complex as that?

Re: Issue with Job executioner removing the job record on server restart

satramalin — Wed, 18 Feb 2015 22:01:59 GMT

Thanks for your response. The last wait state was a timer and the timer got fired which pushed the instance to the service task. The JVM now restarts when the execution is on the service task. If I am not wrong, the execution should now go to the last wait state which is the timer(timer is attached to the user task). This timer has already fired and there are no records in the jobs table. In this case, the instance gets stuck at the system task without moving forward. Attached is the .bpmn for your observation. Can you please help with a new design if this does not suit.

Re: Issue with Job executioner removing the job record on server restart

dmoturu — Wed, 18 Feb 2015 22:58:14 GMT

Hi Joram & other experts,

We have a scenario I think it is same as above.
In our BPD we have a start event –> Service Task1 –> User task (with 2 days due date) with a attached Timer event that fires after 5 mins–> And the flow from Timer goes back to Service Task 1again.

Goal is to make Service Task1 run in an end less loop.

This BPD is working fine and able to run the Service Task1 in endless loop as designed.

However, It is getting into hung state whenever JVM is restarting (due to any other reason) while Service Task1 is in middle of execution.

Any idea why BPD instance is getting hung state instead of restarting from the Service Task1?

Addition info is we set both Asynchronus and exclusive set to true for Service Task1.

Re: Issue with Job executioner removing the job record on server restart

jbarrez — Sat, 28 Feb 2015 08:30:11 GMT

@satramalin :you have async on the service task. The last stable state is the moment just before the service task. That is why you see what you see. The timer fires, and it commits right before the service task due to the async bit.

@dmoturu : What do you mean with endless loop? And endless Java loop? In that case, that's something I wouldn't do, as you are keeping your transaction open for a long time. Most databases dont like that.

Re: Issue with Job executioner removing the job record on server restart

dmoturu — Tue, 03 Mar 2015 16:47:13 GMT

Thanks Joram for response.

What I meant by endless loop is there is going to be only one process instance is active and running for ever and the flow of it comes back to the Service Task repeatedly at every 5 minutes.

User task (with its due date gets set to 2 days every time it is created) is a dummy and does nothing in this case as the timer fires at 5 minutes after its creation and closed by the Timer.

Timer flow goes back to the Service Task so that it gets executed repeatedly at every 5 mins.

Re: Issue with Job executioner removing the job record on server restart

jbarrez — Tue, 03 Mar 2015 17:34:46 GMT

ok, now i follow your explanation. But i can't quite say why the JVM is hanging. Also: how do you verify this? Isn't it the database that is waiting on someting?

The theory of Activiti and transactions is simple: server goes down -> transaction rollback