[ 
https://issues.apache.org/jira/browse/YARN-6094?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15826665#comment-15826665
 ] 

Sangjin Lee commented on YARN-6094:
-----------------------------------

Thanks for the patch [~vrushalic]!

As for the unit test failures, they are related with the changes in the patch. 
{{DataGeneratorForTests}} assumes the presence of the {{timelineservice}} jar 
under its {{target}} directory, but that is true only if {{mvn package}} was 
done on that module. The unit tests don't require that. We need a different way 
to get at the location of the jar (coming from the maven local cache).

Question: I see we're making a change to upload and use the jar that has the 
co-processor from hdfs, rather than adding it to the hbase install. Is that 
change needed for the dynamic co-processor? What is the benefit of doing that 
over using it out of the local install?

(FlowRunCoprocessor.java)
- maybe I'm bit on the paranoid side, but checking whether it is operating on 
the right table seems like a reasonable sanity check nonetheless; do we need to 
remove this check?

(DataGeneratorForTest.java)
- Do we need to move the package for this and {{TestFlowDataGenerator}}? I'm 
curious as to why that move is necessary. The diffs are now deleting the old 
files and creating the new files so it's not so easy to see what has changed.


> Update the coprocessor to be a dynamically loaded one
> -----------------------------------------------------
>
>                 Key: YARN-6094
>                 URL: https://issues.apache.org/jira/browse/YARN-6094
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>            Reporter: Vrushali C
>            Assignee: Vrushali C
>         Attachments: YARN-6094.001.patch, YARN-6094-YARN-5355.001.patch, 
> YARN-6094-YARN-5355.002.patch
>
>
> The timeline service v2 code base on yarn now uses hbase 1.2.4 after 
> YARN-5976. 
> With this version of hbase, system classes (starting with org.apache.hadoop) 
> can be loaded as table level coprocessors. Hence we should update the 
> timeline service coprocessor to be a dynamically loaded one instead of static 
> loading. 
> This involves code changes as well as documentation updates for deployment.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to