[ 
https://issues.apache.org/jira/browse/YARN-10254?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17102593#comment-17102593
 ] 

Gergely Pollak commented on YARN-10254:
---------------------------------------

[~pbacsko] thank you for your feedback,

#1) Yes it got smaller and smaller each iteration I made changes to the class, 
and I just didn't want to remove it yet, but you are right, there is no point 
in keeping this method. Originally I wanted to have some validation here, but 
those were effectively the same as the validations in other methods. Will 
remove it in next patch

#2) It is a good question, I honestly don't know which is the better approach, 
my thought process was if the mapping rule is invalid we still might want to 
place the applciation to default queue rather than bouncing it back, but this 
case is a kind of misconfiguration, where an exception could be in place. On 
the other hand configuration errors should fail at initialization time. So I'm 
not sure which is the best solution, thoghts anyone?

#3) Will be fixed :)

#4) Ditto

#5) Yes you are right, we should query the queue by parent.getQueuePath() + "." 
+ leafQueueName, because the get wont recoginze a short parent name + a short 
leaf name, only a full parent path + short leafName.

#6) Alter mapping just alters a mapping, it depends on the user if it is a 
normalization, or just a simple overwrite, so that's why the method name isn't 
more expressive. comment will be added.

 

> CapacityScheduler incorrect User Group Mapping after leaf queue change
> ----------------------------------------------------------------------
>
>                 Key: YARN-10254
>                 URL: https://issues.apache.org/jira/browse/YARN-10254
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Gergely Pollak
>            Assignee: Gergely Pollak
>            Priority: Major
>         Attachments: YARN-10254.001.patch, YARN-10254.002.patch
>
>
> YARN-9879 and YARN-10198 introduced some major changes to user group mapping, 
> and some of them unfortunately had some negative impact on the way mapping 
> works.
> In some cases incorrect PlacementContexts were created, where full queue path 
> was passed as leaf queue name. This affects how the yarn cli app list 
> displays the queues.
> u:%user:%primary_group.%user mapping fails with an incorrect validation error 
> when the %primary_group parent queue was a managed parent.
> Group based rules in certain cases are mapped to root.[primary_group] rules, 
> loosing the ability to create deeper structures.
>  



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to