[ 
https://issues.apache.org/jira/browse/HBASE-13103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14484842#comment-14484842
 ] 

Mikhail Antonov commented on HBASE-13103:
-----------------------------------------

Thanks for review guys!

bq. I'd like this to eventually be a globally enabled feature, with opt-out via 
table configuration. For it's initial commit, it should probably be opt-in 
instead. Having a global kill switch is probably a good idea too.

How about adding a boolean flag it table descriptor (false by default) to bring 
a table to normalizers attention? Global switch would be still in place to turn 
everything off if so desired (probably just like with balancer, there shoud be 
admin rpc call to turn balancer on/off?)

.bq Yes, priorities will become a useful feature. I think what you have here is 
a nice, committable first pass though.
Thanks, yeah - that would probably be subsequent patch (as well as running in 
parallel on all tables opted-in)

.bq Why no touch regions with fewer than 3 regions?
2 considerations I had in mind - tables with fewer than 3 regions probably too 
small to be hot spots in terms of cluster throughout (that may be not true?), 
and also they would require more thought-out rules. Let's say you have table 
with 2 regions, 10 and 70 gb. This current logic would say - well, 70 is < than 
40 * 2, so no normalization for you :) What you may really want is 2 regions of 
40, or 5 regions by 16. Need to have "ideal" region size? 






> [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