[
https://issues.apache.org/jira/browse/YUNIKORN-2811?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17879783#comment-17879783
]
Yu-Lin Chen commented on YUNIKORN-2811:
---------------------------------------
[~ccondit] Thanks for the feedback. I think we can put Spark App ID to the
white list.
The use case in my mind is someone submit a new jobs and try to override the
SparkOperator’s application name in Gateway or any Customized admission
controller. Take below submitssion flow as example:
# Job Submistion → Gateway A({*}yunikorn.apache.org/app-id{*}) →
SparkOperator(spark-app-selector)
# Job Submistion → Gateway A({-}*applicationId*{-}) →
SparkOperator(spark-app-selector) → Customized Admission Controller
({*}yunikorn.apache.org/app-id{*})
The first one should pass since there is only one appllication ID except Spark
App ID , but the second one should be rejected.
Currently, YuniKorn only supports one third-party label “spark-app-selector”
from SparkOperator, so I think we should only consider it. If user add multiple
metadata in other places (Gateway or AM) , the inconsistencies must be
addressed outside YuniKorn to prevent any confusion.
> Reject task if a pod has inconsistent metadata in the shim
> ----------------------------------------------------------
>
> Key: YUNIKORN-2811
> URL: https://issues.apache.org/jira/browse/YUNIKORN-2811
> Project: Apache YuniKorn
> Issue Type: Sub-task
> Components: shim - kubernetes
> Reporter: Yu-Lin Chen
> Assignee: Yu-Lin Chen
> Priority: Major
> Labels: release-notes
>
> This is a followup of YUNIKORN-2810.
> We should reject the task If the pod metadata (such as queue or
> applicationID) is configured inconsistently.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
