topic Re: Activiti job executors in a clustered environment in Alfresco Archive

Activiti job executors in a clustered environment

peter_spikings — Thu, 21 Jun 2012 10:10:53 GMT

Hi,I've got a two node back-end cluster driven by Activiti (front-ends start processes and the back-ends poll for activities they can process). All good but soon I'm going to be wanting to have timer start events. I'd like to enable the job executor on both but wouldn't want both of the back-ends to

Re: Activiti job executors in a clustered environment

peter_spikings — Mon, 25 Jun 2012 07:03:01 GMT

Bump

Re: Activiti job executors in a clustered environment

ronald_van_kuij — Mon, 25 Jun 2012 07:38:27 GMT

http://www.catb.org/~esr/faqs/smart-questions.html#idp30004672

Re: Activiti job executors in a clustered environment

frederikherema1 — Mon, 25 Jun 2012 09:27:21 GMT

Thanks for the reminder, Ronald.

The job-executors can run concurrently. When they both acquire the same job, one of them will get an optimistic-lock exception when executing the timer and will roll back.

Re: Activiti job executors in a clustered environment

peter_spikings — Mon, 25 Jun 2012 09:36:03 GMT

Thanks for the reminder, Ronald.

The job-executors can run concurrently. When they both acquire the same job, one of them will get an optimistic-lock exception when executing the timer and will roll back.

Excellent, I suspected as much, just wanted to confirm!

Thanks for the help,

Peter.

Re: Activiti job executors in a clustered environment

slash117 — Wed, 11 Sep 2013 16:43:04 GMT

hello,
I have the same situation on my production platform. I was wondering if its not better to use a lock based on the database ("select from 'specificLockTable' for update" in the beginning of the transaction used by the jobExecutor) than to catch the OptimisticLockingFailure Exception. This way we ensure that no concurrent access happens when different jobExecutors are executed.

Regards,
Gilbert E.H.

Re: Activiti job executors in a clustered environment

trademak — Thu, 12 Sep 2013 06:36:20 GMT

Hi,

What's the issue with the current implementation? When you have multiple job executors, only 1 will execute a job as the job is locked in the database. Another job executor will give an OptimisticLockException. So what's wrong with that?

Best regards,