[jira] [Created] (UIMA-3390) DUCC job driver (JD) should be more cautios about queuing work items in the presence of failed ones

["Lou DeGenaro (JIRA)" <de...@uima.apache.org>]

Lou DeGenaro created UIMA-3390:
----------------------------------

             Summary: DUCC job driver (JD) should be more cautios about queuing work items in the presence of failed ones
                 Key: UIMA-3390
                 URL: https://issues.apache.org/jira/browse/UIMA-3390
             Project: UIMA
          Issue Type: Bug
          Components: DUCC
            Reporter: Lou DeGenaro
            Assignee: Lou DeGenaro


I think I understand this problem.  Bhavani's job 96348 happens to have excellent characteristics with 1 JP and 1 thread.

It seems that one of her work items causes the JP to explode (Croaked).  The work items fails.  The JD marks the work item as Error.  So far so good.  The JD proceeds to queue the next work item for processing not realizing that the lone JP has vanished.  This is a mistake.  Correctly, a new JP is allocated by the Scheduler, and not until its initialization is completed does the longtime queued work item get picked up.

The JD should probably be more cautious about queuing work items in the presence of failed ones.



--
This message was sent by Atlassian JIRA
(v6.1#6144)