You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@uima.apache.org by "Lou DeGenaro (JIRA)" <de...@uima.apache.org> on 2013/06/06 20:45:20 UTC
[jira] [Commented] (UIMA-2752) DuccJobSubmit option
--process_failures_limit is ignored
[ https://issues.apache.org/jira/browse/UIMA-2752?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13677360#comment-13677360 ]
Lou DeGenaro commented on UIMA-2752:
------------------------------------
The limit is enforced by the HealthMonitor (HM) which runs according to the following value in ducc.properties: ducc.orchestrator.maintenance.rate=60000
It is entirely possible to have (many) more process failures than the specified limit between(HM) intervals. The only guarantee is that when the HM does run, any job over its failures limit will be killed.
I'm re-assigning back to Burn to either close or advise further.
> DuccJobSubmit option --process_failures_limit is ignored
> --------------------------------------------------------
>
> Key: UIMA-2752
> URL: https://issues.apache.org/jira/browse/UIMA-2752
> Project: UIMA
> Issue Type: Bug
> Components: DUCC
> Environment: When set to 2 the job was killed after 15 work-items failed.
> P7 job 1287 shows 5 failing JPs each with 3 errors. The scaleout was 1 (e.g. all failed in the same thread 71) but since the failures were fast it produced 3 before the job was cancelled. Since there was only 1 thread these extra failures should probably not have been considered collateral damage and retried.
> Is the limit work-item failures or process failures?
> Reporter: Burn Lewis
> Assignee: Lou DeGenaro
> Priority: Minor
> Fix For: 1.0-Ducc
>
>
--
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