[
https://issues.apache.org/jira/browse/TAJO-44?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hyunsik Choi updated TAJO-44:
-----------------------------
Attachment: TAJO-44.patch
I'm still working on this issue. I've upload the first patch for sharing the
progress.
This patch adopts Yarn's AMRMClientImpl to RMContainerAllocator and
RMCommunicator. After this change, RMContainerAllocator and SubQuery become
cleaner and simpler than before. Also, I have never experienced the hanging
problem as I mentioned in TAJO-42.
Besides, this patch cleans up some related things as follows:
* Remove tajo.master.Priority and replace it by
org.apache.hadoop.yarn.api.records.Priority.
* Clean up SubQuery
** Remove a transition from RUNNING to SUCCEEDED and merge it to another
transition.
** add setStartTime() and setFinishTime() methods
** Clean up SubQuery::cleanUp() and rename it to buildAndSetTableMeta().
** Clean up InitAndRequestContainer::transition method.
*** Many common codes are extracted to some methods.
*** The code flow becomes very simple.
> Adopt AMRMClient to RMContainerAllocator, RMCommunicator
> --------------------------------------------------------
>
> Key: TAJO-44
> URL: https://issues.apache.org/jira/browse/TAJO-44
> Project: Tajo
> Issue Type: Improvement
> Reporter: Hyunsik Choi
> Assignee: Hyunsik Choi
> Fix For: 0.2-incubating
>
> Attachments: TAJO-44.patch
>
>
> Hadoop Yarn 2.0.3 introduced AMRMClient that contains common utility methods
> for AMRMProtocol. We can make RMContainerAllocator and RMCommunicator simpler
> by adopting AMRMClient to them.
--
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
