hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Yu Li (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-14456) Implement a namespace-based region grouping strategy for RegionGroupingProvider
Date Tue, 22 Sep 2015 12:35:04 GMT

     [ https://issues.apache.org/jira/browse/HBASE-14456?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Yu Li updated HBASE-14456:
    Attachment: HBASE-14456_v3.patch

Resolves checkstyle issue reported by HadoopQA.

Regarding javadoc warnings, checking the output but find no new warning related to patch of
this JIRA. Zombie tests should be also irrelative.

Ask HadoopQA to re-run.

> Implement a namespace-based region grouping strategy for RegionGroupingProvider
> -------------------------------------------------------------------------------
>                 Key: HBASE-14456
>                 URL: https://issues.apache.org/jira/browse/HBASE-14456
>             Project: HBase
>          Issue Type: New Feature
>    Affects Versions: 2.0.0, 1.3.0
>            Reporter: Yu Li
>            Assignee: Yu Li
>             Fix For: 2.0.0, 1.3.0
>         Attachments: HBASE-14456.patch, HBASE-14456_v2.patch, HBASE-14456_v3.patch
> In HBASE-5699 we introduced multiple wal, and after some succeeding works we now have
two main kinds of wal providers: DefaultWALProvider and RegionGroupingProvider, while DefaultWALProvider
supplies a _single_ wal per regionserver, RegionGroupingProvider supplies _multiple_ wals
according to the RegionGroupingStrategy it's using.
> Now there're two kinds of RegionGroupingStrategy: _identity_ and _bounded_, in which
"identity" only for testing (trial) purpose and "bounded" for randomly assigning region writes
to bounded number of wals. Although the "bounded" strategy is good enough for IO leverage,
we may still want a more "designed" way to distribute the region writes, like having one wal
group per business.
> Since we already have namespace for multi-tenancy, it would be good to have namespace-based
grouping strategy, and this is exactly what this JIRA is for.
> There might be more benefits if we have namespace-based wal group, like only replicating
data for the namespace in need w/o filtering anything useless, or only force region flush
for the heavy loaded namespace, etc. although we  won't include all these in this single JIRA.

This message was sent by Atlassian JIRA

View raw message