[
https://issues.apache.org/jira/browse/HBASE-10576?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13980572#comment-13980572
]
Jeffrey Zhong commented on HBASE-10576:
---------------------------------------
AM & Load Balancers do have a similar handling for regions of a "disabled"
table. The proposal is suggesting a way to have a stronger guarantee of the
region collocating as you know AM doesn't guarantee when an assignment will
happen and the final location of a region assignment. For example, when an
index region move, there is no guarantee that data region is also moved and
opened simultaneously on the same RS. It's possible that during an index update
a region move could happen and the update survives the region move because of
retires. So there are chances index region & data region on different RSs even
within one update operation. This may cause client gets a fail response(like
time out or interrupted) while the data is really committed and data
region(vice verse) may don't commit the related updates and finally end up with
index out of sync with data.
If current implementation can handle those similar corner cases from
indeterministic nature of AM, I think current load balancer approach is also
good.
> Custom load balancer to co-locate the regions of two tables which are having
> same split keys
> --------------------------------------------------------------------------------------------
>
> Key: HBASE-10576
> URL: https://issues.apache.org/jira/browse/HBASE-10576
> Project: HBase
> Issue Type: Sub-task
> Components: Balancer
> Reporter: rajeshbabu
> Assignee: rajeshbabu
> Attachments: HBASE-10536_v2.patch, HBASE-10576.patch
>
>
> To support local indexing both user table and index table should have same
> split keys. This issue to provide custom balancer to colocate the regions of
> two tables which are having same split keys.
> This helps in Phoenix as well.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
