[ 
https://issues.apache.org/jira/browse/HBASE-23006?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16926672#comment-16926672
 ] 

Duo Zhang commented on HBASE-23006:
-----------------------------------

The fix is simple, just removed lots of code in RSGroupBasedLoadBalancer... Not 
sure why we introduce lots of code to deal with misplaced regions specially in 
retainAssignment, just give the internal load balancer the correct candidate 
list, if the region is misplaced, it will be place to new places...

> RSGroupBasedLoadBalancer should also try to place replicas for the same 
> region to different region servers
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-23006
>                 URL: https://issues.apache.org/jira/browse/HBASE-23006
>             Project: HBase
>          Issue Type: Bug
>          Components: Region Assignment, rsgroup
>            Reporter: Duo Zhang
>            Assignee: Duo Zhang
>            Priority: Major
>         Attachments: HBASE-23006-UT.patch
>
>
> Found this when implementing HBASE-22971. TestSCPWithReplicas fails when 
> RSGroupBasedLoadBalancer is enabled.
> And this can be reproduced by a UT on master branch too. I think the problem 
> is that in RSGroupBasedLoadBalancer.retainAssignment we do not consider 
> region replicas.
> We should fix this.



--
This message was sent by Atlassian Jira
(v8.3.2#803003)

Reply via email to