hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-16803) Make hbase:acl table unsplittable
Date Tue, 11 Oct 2016 20:23:20 GMT

    [ https://issues.apache.org/jira/browse/HBASE-16803?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15566474#comment-15566474
] 

stack commented on HBASE-16803:
-------------------------------

HBASE-11165 is the scaling issue w/ attached issues and discussion. "[DISCUSS] No regions
on Master node in 2.0" is one thread that talks about this topic with attached docs that discuss
pros and cons. Y! production is running a hacked hbase that splits hbase:meta. They are trying
to get the facility back upstream ([~toffer] FYI). The new AM is doing its best to support
splittable hbase:meta. More work to do. A patch that makes an argument that hbase:acl should
not split because hbase:meta does not split is specious and going in the wrong direction.

> Make hbase:acl table unsplittable
> ---------------------------------
>
>                 Key: HBASE-16803
>                 URL: https://issues.apache.org/jira/browse/HBASE-16803
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: Ted Yu
>            Assignee: Ted Yu
>         Attachments: 16803.v1.txt, 16803.v2.txt
>
>
> HBASE-16773 fixed a case where PriorityRpcServer handler threads are all occupied accessing
hbase:acl table.
> However, the fix relies on the fact that there is single region in hbase:acl table so
that the access can be local.
> As discussed at the end of HBASE-16773, we should disable split of hbase:acl table as
well.
> hbase:meta is normally much larger than hbase:acl table and it has only one region.



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

Mime
View raw message