[jira] Updated: (UIMA-1165) Fix a race condition in the UIMA-AS aggregate that results in a hang

["Jerry Cwiklik (JIRA)" <ui...@incubator.apache.org>]

     [ https://issues.apache.org/jira/browse/UIMA-1165?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jerry Cwiklik updated UIMA-1165:
--------------------------------

    Attachment: uimaj-as-jms-UIMA-1165-patch.txt
                uimaj-as-core-UIMA-1165-patch.txt
                uimaj-as-activemq-UIMA-1165-patch.txt

Synchronized critical sections in the UIMA AS Aggregate to fix a race condition that lead to a hang. Modified class boolean global vars to be volatile to make sure the changes are immediately visible to all the threads. Made several changes to remove inefficient synchronization using InProcessCache object which effectively locked that entire object. Removed dead code. 

> Fix a race condition in the UIMA-AS aggregate that results in a hang
> --------------------------------------------------------------------
>
>                 Key: UIMA-1165
>                 URL: https://issues.apache.org/jira/browse/UIMA-1165
>             Project: UIMA
>          Issue Type: Bug
>          Components: Async Scaleout
>            Reporter: Jerry Cwiklik
>         Attachments: uimaj-as-activemq-UIMA-1165-patch.txt, uimaj-as-core-UIMA-1165-patch.txt, uimaj-as-jms-UIMA-1165-patch.txt
>
>
> When running a complex Cas Multiplier Aggregate AS service with a scaled out listener threads, a race condition occurs that leads to a hang. Specifically, one thread obtains a number of child CASes produced from an input CAS, than this thread is preempted and another thread releases one of the child CASes. The parent CAS may not be released from the InProcessCache leading to CAS pool exhaustion since many CASes are stuck in the InProcessCache waiting to be released.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.