[
https://issues.apache.org/jira/browse/HBASE-25517?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Xiaolin Ha updated HBASE-25517:
-------------------------------
Description:
When we call balanceRSGroup(), it waits for the lock of balancer, which may be
occupied by the overall cluster balance(). Sometimes the balance of cluster
will take a long time because it need to compute the costs or act the balance
region plans.
In RSGroup, we don't need to block a balance of one group due to other group
balances.
All group are mini clusters, and the balance of their regions should support
concurrently execution and reduce the race conditions between them.
> Reduce race conditions between balanceRSGroup and balance() cluster, and
> support concurrently balance RSGroups
> --------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-25517
> URL: https://issues.apache.org/jira/browse/HBASE-25517
> Project: HBase
> Issue Type: Improvement
> Components: Balancer
> Reporter: Xiaolin Ha
> Assignee: Xiaolin Ha
> Priority: Major
>
> When we call balanceRSGroup(), it waits for the lock of balancer, which may
> be occupied by the overall cluster balance(). Sometimes the balance of
> cluster will take a long time because it need to compute the costs or act the
> balance region plans.
> In RSGroup, we don't need to block a balance of one group due to other group
> balances.
> All group are mini clusters, and the balance of their regions should support
> concurrently execution and reduce the race conditions between them.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
