[
https://issues.apache.org/jira/browse/HBASE-13103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14364754#comment-14364754
]
Mikhail Antonov commented on HBASE-13103:
-----------------------------------------
Some notes as I'm sketching the draft:
- generally looks like the basic balancing architecture would just work - a
chore on master, interface + pluggable initialization, main invocation gets
kicked in HMaster, period and cutoff time limits.
- runs on per-table basis (for first cut - could just do all or nothing, then
add normalization params to table level configuration if needed)
- normalizer computes list of normalization plans (which are simply, either
split R1, or merge R1 and R1), those plans then executed one by one, I guess we
don't want more than one merge or split going on the table in most cases?
Execution of plan is simply figuring currently assigned HRS and requesting
split over rpc.
- whole thing is stateless, if master crashed during normalization, on next
scheduled iteration it will be recomputed anyway
[~ndimiduk] thoughts?
> [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
>
> 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)
