hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Enis Soztutar (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-10351) LoadBalancer changes for supporting region replicas
Date Tue, 25 Feb 2014 20:28:22 GMT

    [ https://issues.apache.org/jira/browse/HBASE-10351?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13912015#comment-13912015

Enis Soztutar commented on HBASE-10351:

bq. Left some comments yesterday and today
Thanks! v4 patch addressed all the reviews there. 
bq. although the array manipulations are rather complex and meaning is not always clear 
When we unwound hashmap + hashset operations into array and index based ones for performance,
we ended up un-javaifying that code. Not much to do about it (see HBASE-8119 for ref) 
bq. The one issue that is there I think is that the test (TestMasterOperationsForReplicas)
may be flaky
It seems fine in my runs. retain assignment now also looks for availability and finds the
best server while retaining the host locality. We can fix it if we see flakiness. 
bq. The comment in the catch-all code should be updated.

> LoadBalancer changes for supporting region replicas
> ---------------------------------------------------
>                 Key: HBASE-10351
>                 URL: https://issues.apache.org/jira/browse/HBASE-10351
>             Project: HBase
>          Issue Type: Sub-task
>          Components: master
>    Affects Versions: 0.99.0
>            Reporter: Enis Soztutar
>            Assignee: Enis Soztutar
>         Attachments: hbase-10351_v0.patch, hbase-10351_v1.patch, hbase-10351_v3.patch
> LoadBalancer has to be aware of and enforce placement of region replicas so that the
replicas are not co-hosted in the same server, host or rack. This will ensure that the region
is highly available during process / host / rack failover. 

This message was sent by Atlassian JIRA

View raw message