[ 
https://issues.apache.org/jira/browse/HBASE-13103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14486812#comment-14486812
 ] 

Lars Hofhansl commented on HBASE-13103:
---------------------------------------

Just reading through the comments here. Unless the "reshaper" perfectly takes 
all factors into account I'd be very hesitant to run it on our clusters. With 
perfect I mean that it knows about load, disk IO, etc. Since that's hard (or 
impossible) I think I'd prefer to trigger this manually, as suggested in the 
description. But maybe I am overly cautious.

Split decisions can be made locally and are rarely bad (unless really 
excessive). Merge decisions need (a) global knowledge - not all regions may on 
the same server and (b) can possibly lead to worse performance (hot regions 
merged together, etc)
What we can automate is merging empty regions away.


> [ergonomics] add region size balancing as a feature of master
> -------------------------------------------------------------
>
>                 Key: HBASE-13103
>                 URL: https://issues.apache.org/jira/browse/HBASE-13103
>             Project: HBase
>          Issue Type: Brainstorming
>          Components: Usability
>            Reporter: Nick Dimiduk
>            Assignee: Mikhail Antonov
>             Fix For: 2.0.0, 1.1.0
>
>         Attachments: HBASE-13103-v0.patch
>
>
> Often enough, folks miss-judge split points or otherwise end up with a 
> suboptimal number of regions. We should have an automated, reliable way to 
> "reshape" or "balance" a table's region boundaries. This would be for tables 
> that contain existing data. This might look like:
> {noformat}
> Admin#reshapeTable(TableName, int numSplits);
> {noformat}
> or from the shell:
> {noformat}
> > reshape TABLE, numSplits
> {noformat}
> Better still would be to have a maintenance process, similar to the existing 
> Balancer that runs AssignmentManager on an interval, to run the above 
> "reshape" operation on an interval. That way, the cluster will automatically 
> self-correct toward a desirable state.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to