Mailing-List: contact issues-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Date: Tue, 20 Jan 2015 18:15:35 +0000 (UTC)
From: "Andrey Stepachev (JIRA)" <jira@apache.org>
To: issues@hbase.apache.org
Message-ID: <JIRA.12768550.1421706961000.125744.1421777735566@Atlassian.JIRA>
In-Reply-To: <JIRA.12768550.1421706961000@Atlassian.JIRA>
References: <JIRA.12768550.1421706961000@Atlassian.JIRA>
 <JIRA.12768550.1421706961930@arcas>
Subject: [jira] [Commented] (HBASE-12880) RegionState in state SPLIT doesn't
 removed from region states
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HBASE-12880?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14284142#comment-14284142 ] 

Andrey Stepachev commented on HBASE-12880:
------------------------------------------

Yeah. Seems it explicitly tested, that region kept in state map (MERGE/SPLIT), but I have no idea when it should be removed from states map.

[~jxiang], can you take a look. What was an idea to keep regions in MERGE/SPLIT states in states map and where them should be removed?
I fixed that with this patch, but here is tests that explicitly test that we don't allow to assign/unassign merged/split regions. But how long it restricts and when this records will be removed from maps?

Thank you.

> RegionState in state SPLIT doesn't removed from region states
> -------------------------------------------------------------
>
>                 Key: HBASE-12880
>                 URL: https://issues.apache.org/jira/browse/HBASE-12880
>             Project: HBase
>          Issue Type: Bug
>          Components: Region Assignment
>    Affects Versions: 2.0.0, 1.1.0
>            Reporter: Andrey Stepachev
>            Assignee: Andrey Stepachev
>         Attachments: HBASE-12880.patch, master-with-split-regions-2-1.jpg
>
>
> During my work on patch HBASE-7332 I stumbled on strange behaviour in RegionStates. Split region doesn't removed from regionStates in regionOffline() method and RegionState for this region sits in regionStates map indefinitely long (until RS rebooted).
> (that is clearly seen in HBASE-7332 by simple creating table and splitting it from command line).
> Is that was intended to be so and some chore eventually will remove it from regionStates (didn't find with fast code scanning) or here can be resource leak?


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