[jira] [Commented] (OOZIE-2650) Retry coord start on database exceptions

[Abhishek Bafna (JIRA)]

    [ 
https://issues.apache.org/jira/browse/OOZIE-2650?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15436646#comment-15436646
 ] 

Abhishek Bafna commented on OOZIE-2650:
---------------------------------------

I think we can define a new error code instead of using {{E1304, "Duplicate 
bundle application coordinator name"}}

{code}
    @Override
    protected void verifyPrecondition() throws CommandException {
        super.verifyPrecondition();
        if (wfId != null) {
            LOG.warn("Workflow is already submitted for Coord Action [{0}]", 
getParentId());
            throw new CommandException(ErrorCode.E1304, getParentId());
        }
    }
{code}

> Retry coord start on database exceptions
> ----------------------------------------
>
>                 Key: OOZIE-2650
>                 URL: https://issues.apache.org/jira/browse/OOZIE-2650
>             Project: Oozie
>          Issue Type: Bug
>            Reporter: Satish Subhashrao Saley
>            Assignee: Satish Subhashrao Saley
>         Attachments: OOZIE-2650-1.patch
>
>
> {code:title=CoordActionStartXCommand.java}
> updateList.add(new UpdateEntry<WorkflowJobQuery>(
>                             WorkflowJobQuery.UPDATE_WORKFLOW_PARENT_MODIFIED, 
> wfJob));
>                     updateList.add(new UpdateEntry<CoordActionQuery>(
>                             CoordActionQuery.UPDATE_COORD_ACTION_FOR_START, 
> coordAction));
>                     try {
>                         executor.executeBatchInsertUpdateDelete(insertList, 
> updateList, null);
>                         queue(new 
> CoordActionNotificationXCommand(coordAction), 100);
>                         if (EventHandlerService.isEnabled()) {
>                             generateEvent(coordAction, user, appName, 
> wfJob.getStartTime());
>                         }
>                     }
>                     catch (JPAExecutorException je) {
>                         throw new CommandException(je);
>                     }
> .........
> .......
> ........
>  finally {
>                 if (makeFail == true) { // No DB exception occurs
>                     ....
>                     ....
>                     ....
>                     queue(new 
> CoordActionReadyXCommand(coordAction.getJobId()));
>                 }
>             }
> {code}
> If there is any Database issue while starting coord action, we fail the coord 
> action. We should retry. 
> CoordActionStartXCommand submits workflow. Workflow gets linked to the coord 
> action if workflow submission succeeds. But if coord action update fails due 
> to database exception, recovery service should be able to recover it.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)