[jira] [Commented] (FALCON-1572) Only one instance is running in a process when run using Native Scheduler

["Pallavi Rao (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FALCON-1572?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15052495#comment-15052495 ] 

Pallavi Rao commented on FALCON-1572:
-------------------------------------

The issue apart from what Pavan has stated above was that, a list of awaitingInstances was maintained SchedulerService. The service would pick an awaitingInstance from this list and schedule it when a previous instance completed. The problem was since it was a list, whenever an instance was updated (with predicate to wait for job completion) and added to it, it would be treated as a new element. So, there would be 2 elements for the same instance, one with predicate and one without. Since the one without predicate appeared first on the list, the validation for existence of the predicate would fail and no instances would get scheduled. 

Changed the data structure to a SortedMap to allow for updates and also to ensure the instances are picked up based on instanceSequence number.

> Only one instance is running in a process when run using Native Scheduler
> -------------------------------------------------------------------------
>
>                 Key: FALCON-1572
>                 URL: https://issues.apache.org/jira/browse/FALCON-1572
>             Project: Falcon
>          Issue Type: Bug
>            Reporter: pavan kumar kolamuri
>            Assignee: Pallavi Rao
>            Priority: Blocker
>             Fix For: 0.9
>
>         Attachments: FALCON-1572.patch
>
>
> When you submit and schedule a process , only one instance is running and remaining instances are not getting scheduled . Reason because when job completion event comes to falcon, the execution order of services is not taken care. Scheduling Service should run only after Execution Service. As Execution service didn't run state of first instance is not getting updated and other instances scheduling conditions are not met because of this. 



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