jack patwork commented on AMQ-5129:

I've seen the same thing on 5.9.1. I'm running a 4 node cluster and 
intermittently, every few weeks, the scheduler stops working. I haven't yet 
been able to spot what triggers this. I'm using the purgeInactiveDestinations 
on dynamic topics that are being getting and purged a few times a minute so in 
order to keep the memory under control this is critical. When it stops within a 
few hours the affected instance is out of memory and needs a restart.

I think the scheduler definitely needs to be more robust and handle failures 
without stopping. It's a concern that the the scheduler can stop without any 
trace of failure in the logs, apart from a canary watch of the scheduler log 

> Substitute TimeTask with ScheduledExecutorService in 
> org.apache.activemq.thread.Scheduler
> -----------------------------------------------------------------------------------------
>                 Key: AMQ-5129
>                 URL: https://issues.apache.org/jira/browse/AMQ-5129
>             Project: ActiveMQ
>          Issue Type: Wish
>    Affects Versions: 5.9.0
>            Reporter: matteo rulli
>         Attachments: proposed-patch.txt
> Since Timer has only one execution thread, long-running task can affect other 
> scheduled tasks. Besides, runtime exceptions thrown in TimerTasks kill the 
> only one running thread, bringing down the entire Scheduler.
> I have the suspect that all this could relate to AMQ-3938: sometimes in very 
> busy environments I experience exaclty the same problem: a slow leakeage due 
> to temp queues that are not deleted. Since 
> org.apache.activemq.broker.region.RegionBroker uses a Scheduler to activate 
> the purgeInactiveDestinations, a crashed timer could explain why 
> purgeInactiveDestinations stops working.
> I attached a tentative patch to migrate timer to ScheduledExecutorService. 
> Hope this helps.

This message was sent by Atlassian JIRA

Reply via email to