[
https://issues.apache.org/jira/browse/KYLIN-1241?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15069103#comment-15069103
]
Shaofeng SHI commented on KYLIN-1241:
-------------------------------------
Dong, I can understand, this is a good solution and can solve your pain
quickly. While from Kylin's perspective, we hope to provide the flexibility and
keep the usability: it should be easy to use for both user and kylin admin. So
I'm wondering if there is a better way to solve this.
Dividing the job configs by cube capacity into small, medium and large ones was
a historic decision. From my experience it is not very useful; each project may
have both small and large cubes, they have to share the same properties, this
is not flexible.
So, maybe we can allow user to upload a job conf file per project/cube, which
is in the same format as kylin_job_conf.xml. The content will be persisted into
metadata store, and user can upload/delete it from Kylin UI.
When building a cube, Kylin will seek whether there is a cube specific conf
file; if yes will use it; Otherwise will seek the project specific one; Finally
if both cube and project don't have, will use the one that provided by admin in
KYLIN_HOME/conf/.
> support machine resource isolation in project level with different
> kylin_job_conf.xml file
> ------------------------------------------------------------------------------------------
>
> Key: KYLIN-1241
> URL: https://issues.apache.org/jira/browse/KYLIN-1241
> Project: Kylin
> Issue Type: New Feature
> Components: Job Engine
> Affects Versions: v1.1.1
> Reporter: Wang,Dong
> Assignee: ZhouQianhao
> Priority: Minor
> Fix For: v1.3
>
> Attachments: KYLIN-1241.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
