Lou DeGenaro created UIMA-2689:
----------------------------------

             Summary: DUCC webserver (WS) auto-cancel feature should not be 
enabled for "shadows"
                 Key: UIMA-2689
                 URL: https://issues.apache.org/jira/browse/UIMA-2689
             Project: UIMA
          Issue Type: Bug
            Reporter: Lou DeGenaro
            Assignee: Lou DeGenaro
            Priority: Minor


The auto-cancel feature is enabled when --cancel_on_interrupt flag is specified 
at submit time.  When this flag is specified, the WS expects to receive regular 
pings from the submitter, which are automatically supplied by the associated 
monitor.  If pings are absent for too long (config'd in ducc.properties via 
ducc.ws.job.automatic.cancel.minutes) the WS proceeds to cancel the job.

It is possible to run multiple WS instances (aka "shadows").  Since only one is 
ping'd, the other(s) will detect a ping timeout and incorrectly cancel the job.

The solution is to have only the "primary" WS as the monitor+canceler.  This is 
determined by matching the specification in ducc.properties for WS host:port 
with that of the actual host:port of the WS.  Only if these match will the 
auto-cancel code be enabled in the WS.

To bypass this problem until fix is delivered, run only a single WS.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to