[
https://issues.apache.org/jira/browse/UIMA-3577?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13903997#comment-13903997
]
Lou DeGenaro commented on UIMA-3577:
------------------------------------
1. Add constructor to CallbackState
2. Create CasLimbo look-aside buffer comprising those CAS's that are to be
delayed retry
3. Create CasLimbo release() method to enable transfer of delayed CAS's in
look-aside buffer back to active status
4. Create CasSource clone() method to employ CasCopier to copy CAS from
nonreusable CAS to a brand new one
5. Create CasSource release() to release delayed CAS's in limbo
6. Employ CasSource clone() method when fetching CAS;s from limbo
7. Support delayed CAS's in CasTuple
8. Improved FaultInjector, off by default, driven by ducc.properties
ducc.jd.fault.injector.CallBack1 = n, where 0 < n < 1 is the percentage of
callback #1's randomly discarded, 0.1 == 10%
ducc.jd.fault.injector.CallBack2 = n, where 0 < n < 1 is the percentage of
callback #2's randomly discarded, 0.1 == 10%
9. WorkItemListener employ FaultInjector when so configured in ducc.properties
10. In transport in DriverStatusReport, add various methods for workItemsLost
and adjust calculateState() method to account for Lost as being retryable (not
a final state)
> DUCC job driver (JD) should auto-retry lost work items
> ------------------------------------------------------
>
> Key: UIMA-3577
> URL: https://issues.apache.org/jira/browse/UIMA-3577
> Project: UIMA
> Issue Type: Improvement
> Components: DUCC
> Environment: Currently work item is "lost" after 5 minutes and
> callback #2 has not triggered.
> Lost work items are presently not auto-retried.
> Reporter: Lou DeGenaro
> Assignee: Lou DeGenaro
> Fix For: 2.0-Ducc
>
>
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
