Mailing-List: contact commits-help@helix.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@helix.incubator.apache.org
Date: Mon, 28 Oct 2013 01:18:30 +0000 (UTC)
From: "Kanak Biscuitwala (JIRA)" <jira@apache.org>
To: commits@helix.incubator.apache.org
Message-ID: <JIRA.12675357.1382570170153.9330.1382923110388@arcas>
In-Reply-To: <JIRA.12675357.1382570170153@arcas>
References: <JIRA.12675357.1382570170153@arcas>
Subject: [jira] [Resolved] (HELIX-277) FULL_AUTO rebalancer should not
 prefer nodes that are just coming up
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/HELIX-277?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Kanak Biscuitwala resolved HELIX-277.
-------------------------------------

    Resolution: Fixed

Updated the algorithm and added a test. Code review: https://reviews.apache.org/r/14895/

> FULL_AUTO rebalancer should not prefer nodes that are just coming up
> --------------------------------------------------------------------
>
>                 Key: HELIX-277
>                 URL: https://issues.apache.org/jira/browse/HELIX-277
>             Project: Apache Helix
>          Issue Type: Sub-task
>            Reporter: Kanak Biscuitwala
>            Assignee: Kanak Biscuitwala
>
> Rebalancing in FULL_AUTO mode will relentlessly (though mostly greedily) seek a balance in both partitions and states across nodes with minimal movement. However, in this pursuit, we may end up requiring multiple transitions to get a node in the right state for a given partition.
> We can fix this by refusing to prefer nodes as they come up, and then revert to the existing strategy once their current state contains a replica.


--
This message was sent by Atlassian JIRA
(v6.1#6144)