[
https://issues.apache.org/jira/browse/HBASE-2480?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12860029#action_12860029
]
stack commented on HBASE-2480:
------------------------------
Load balancing is also currently too aggressive. It cannot run while a server
crash is being processed. Region splits should not put two daughters back on
the parent's RegionServer (often the balancer will come along soon after and
move one of the just opened daughters ... sort ofhbase-2240).
> Smarter load-balancing
> ----------------------
>
> Key: HBASE-2480
> URL: https://issues.apache.org/jira/browse/HBASE-2480
> Project: Hadoop HBase
> Issue Type: Improvement
> Reporter: stack
>
> Balancer is really basic; it just looks at counts of regions at the moment.
> It also as it currently works has loads of issues (See the linked issues)
> including a lack of testibility. This issue acts as the umbrella issue to
> collect all balancer improvements under.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
