[
https://issues.apache.org/jira/browse/HBASE-21820?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16777482#comment-16777482
]
Guanghao Zhang commented on HBASE-21820:
----------------------------------------
Here is just a simple implemention of CLUSTER quota. As the throttling job is
done by every RS, RS need to know how many quota it can use. Here use a simple
way for this. Every RS use CLUSTER quota * / RS num as it's quota to throttle.
It is not precise when there are hot RS. But I thought we can implement a
better solution in the future. Maybe add a control center in master and
allocate useful quota to every RS.
> Implement CLUSTER quota scope
> -----------------------------
>
> Key: HBASE-21820
> URL: https://issues.apache.org/jira/browse/HBASE-21820
> Project: HBase
> Issue Type: Sub-task
> Reporter: Yi Mei
> Assignee: Yi Mei
> Priority: Major
> Attachments: HBASE-21820.master.001.patch,
> HBASE-21820.master.002.patch, HBASE-21820.master.003.patch,
> HBASE-21820.master.004.patch, HBASE-21820.master.005.patch
>
>
> There are two kinds of quota scope: CLUSTER and MACHINE. CLUSTER quota means
> quota limit is shared by all machines of cluster. MACHINE quota means quota
> limit is used by single region server.
> Currently, all set quota operations use MACHINE scope as default and CLUSTER
> scope has not been implemented. So open this issue to implement CLUSTER quota
> scope.
> To split cluster quota limit to machines, the basic idea is for user,
> namespace, user over namespace and region server quota, use [ClusterLimit /
> RSNum] as machine limit. For table and user over table quota, use
> [ClusterLimit / TotalTableRegionNum * MachineTableRegionNum] as machine
> limit. Suggestions are welcomed.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
