topic Re: Parallel getway fork can't execute task service concurrently in Alfresco Archive

Parallel getway fork can't execute task service concurrently

craig_wu9 — Tue, 24 Jul 2012 02:21:13 GMT

I got tow task service named task1 and task2, both with activiti:async="true" set. And I have active the JobExecutor at activiti.cfg.xml. But the result is the two tasks executed sequence.It will print task1:1 to task1:49 and pause, then continue to print task1: 50 to task1: 100. After that start to

Re: Parallel getway fork can't execute task service concurrently

frederikherema1 — Tue, 24 Jul 2012 07:15:10 GMT

How do you "know" they are run sequential? The first task will probably be on the queue first, and instantly waits for one second (1000 millis). Have you tried with a longer sleep-time? Does task 2 still gets executed after task1?

If so, double-check you're using an activiti version that support async (I think 5.8 or 5.9). If this doesn't do the trick, debug some more and see if the "executeJobs" method is called twice and check what pool-size is used…

Re: Parallel getway fork can't execute task service concurrently

craig_wu9 — Tue, 24 Jul 2012 08:22:45 GMT

Thanks for your replay.

The version of the activiti I used is 5.9.

I change the task 1 to sleep 10 seconds with Thread.sleep(10000), the result is same.

The method "executeJobs" is called twice.
First time, the parameter jobIds has tow elements (I mentioned before, the tow elements are job id of the task1 and task2 in table act_ru_job).
Second time, the parameter jobIds is an empty list. And when I pause at here(breakpoint), I can see the output of task1 and task2.
This means, when the first time call the "executeJobs", the tow tasks are all executed.

I checked the "threadPoolExecutor" the size is default setting. CoreSize =3, maximumPoolSize=10

Re: Parallel getway fork can't execute task service concurrently

frederikherema1 — Tue, 24 Jul 2012 09:29:56 GMT

Didn't get that the first time that it was TWO (not tow) id's passed in the list. So it makes sense they are executed sequentially, the question is why they are offered to the job-executor in batch instead of each job at a time.

See http://activiti.org/userguide/index.html#exclusiveJobs > jobs in one process-instance are by default exclusive, read the section there to get an insight of why this is

Re: Parallel getway fork can't execute task service concurrently

ronald_van_kuij — Tue, 24 Jul 2012 09:51:15 GMT

Didn't Daniel Meyer implement an 'optimistic-locking-exception-job-executer-retry-to-commit-only' mechanism to circumvent the optimistic locking thing written in the 'why exclusive jobs' part?

http://jira.codehaus.org/browse/ACT-436

yes, I know Activiti is not a multi-threading system, but it limits functionality in that you have to start sending messages to e.g. external jms if you want to have concurrency or???

Re: Parallel getway fork can't execute task service concurrently

frederikherema1 — Tue, 24 Jul 2012 11:16:27 GMT

Sure, you can have them execute in parallel, just turn the "exclusive" to false on those async-tasks in the same process.

Re: Parallel getway fork can't execute task service concurrently

ronald_van_kuij — Tue, 24 Jul 2012 13:28:37 GMT

Yes, I know that, but to circumvent troubles due to the OLE, you could configure this 'retry' mechnism, right? (And I do not mean the normal job retry mechanism)

Re: Parallel getway fork can't execute task service concurrently

frederikherema1 — Tue, 24 Jul 2012 13:30:04 GMT

Ronald, I'm not aware of this kind of mechanism being in place, but you can be right though…

Re: Parallel getway fork can't execute task service concurrently

craig_wu9 — Wed, 25 Jul 2012 07:29:32 GMT

Thanks for your replay.
I got what I want.
But also got an exception

org.activiti.engine.ActivitiOptimisticLockingException: ExecutionEntity[6201] was updated by another transaction concurrently

Re: Parallel getway fork can't execute task service concurrently

ronald_van_kuij — Wed, 25 Jul 2012 19:10:57 GMT

Ronald, I'm not aware of this kind of mechanism being in place, but you can be right though…

I'll try to find it… and let you know.