[
https://issues.apache.org/jira/browse/HBASE-13964?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14606687#comment-14606687
]
Nick Dimiduk commented on HBASE-13964:
--------------------------------------
Quota support was apparently added with HBASE-8410. I don't know the reason for
enforcing a number of regions, through I assume this is because a multi-tenent
cluster with a single namespace per user/group would not was a single namespace
consuming all available region count capacity. Maybe [~toffer] can fill us in
on this?
That said, I think we can leave normalization enabled even when quotas are in
place. The normalizer will work toward the optimal physical partitioning of
data until it reaches the limits enforced by the quota policies. After this
point, it will not be able to make any further progress when that progress
requires splitting regions. I think it's the job of the normalization strategy
to take quota restrictions into account. IMHO, a well-behaved normalizer would
remove the need for namespace region # quotas entirely.
I don't really follow why this change is necessary.
> Skip region normalization for tables under namespace quota
> ----------------------------------------------------------
>
> Key: HBASE-13964
> URL: https://issues.apache.org/jira/browse/HBASE-13964
> Project: HBase
> Issue Type: Task
> Components: Balancer, Usability
> Reporter: Mikhail Antonov
> Assignee: Ted Yu
> Fix For: 2.0.0, 1.2.0, 1.3.0
>
> Attachments: 13964-branch-1-v2.txt, 13964-branch-1-v3.txt,
> 13964-v1.txt
>
>
> As [[email protected]] pointed out in HBASE-13103, we need to discuss how to
> normalize regions of tables under namespace control. What was proposed is to
> disable normalization of such tables.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
