Mailing-List: contact hbase-dev-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hbase-dev@hadoop.apache.org
Message-ID: <771123019.1245711607454.JavaMail.jira@brutus>
Date: Mon, 22 Jun 2009 16:00:07 -0700 (PDT)
From: "stack (JIRA)" <jira@apache.org>
To: hbase-dev@hadoop.apache.org
Subject: [jira] Commented: (HBASE-1508) Shell "close_region" reveals a
 Master<>HRS problem, regions are not reassigned
In-Reply-To: <131775024.1244622427476.JavaMail.jira@brutus>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HBASE-1508?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12722882#action_12722882 ] 

stack commented on HBASE-1508:
------------------------------

I broke close_region with this commit:

{code}
------------------------------------------------------------------------
r773168 | stack | 2009-05-09 06:08:58 +0000 (Sat, 09 May 2009) | 1 line

HBASE-1344  WARN IllegalStateException: Cannot set a region as open if it has not been pending -- part2
{code}

Its the addition of this bit of code that is problem (around #355):

{code}
      if (info.isOffline() || this.master.regionManager.regionIsInTransition(info.getRegionNameAsString()) ||
          (serverName != null && this.master.serverManager.isDead(serverName))) {
        return;
      }
{code}

We added above checks so catalog scanners would not fight the shutdown scanner work.  The above is overly constraining.

Changing it to this makes close_region for both binary and non-binary keys work again:

{code}
      if (info.isOffline() ||
        (serverName != null && this.master.regionManager.regionIsInTransition(info.getRegionNameAsString())) ||
          (serverName != null && this.master.serverManager.isDead(serverName))) {
        return;
      }
{code}

Here, if serverName is null, then we've not assigned OR its been closed by somethign like 'close_region' (which clears the start code and server from .META. for the particular region).

> Shell "close_region" reveals a Master<>HRS problem, regions are not reassigned
> ------------------------------------------------------------------------------
>
>                 Key: HBASE-1508
>                 URL: https://issues.apache.org/jira/browse/HBASE-1508
>             Project: Hadoop HBase
>          Issue Type: Bug
>          Components: master
>    Affects Versions: 0.20.0
>            Reporter: Lars George
>            Priority: Blocker
>             Fix For: 0.20.0
>
>
> When issuing a "close_region" on the shell the Master logs these entries:
> {code}
> ...
> 2009-06-09 22:11:31,141 DEBUG org.apache.hadoop.hbase.master.RegionManager: Applying operation in tasklists to region
> 2009-06-09 22:11:33,557 DEBUG org.apache.hadoop.hbase.master.HMaster: Attempting to close region: TestTable,0000291328,1244572849139
> 2009-06-09 22:11:33,560 INFO org.apache.hadoop.hbase.master.HMaster: Marking TestTable,0000291328,1244572849139 as closed on 192.168.2.103:63745; cleaning SERVER + STARTCODE; master will tell regionserver to close region on next heartbeat
> 2009-06-09 22:11:34,156 DEBUG org.apache.hadoop.hbase.master.RegionManager: Applying operation in tasklists to region
> ...
> {code}
> But that is it, no further processing is done. The regions stays closed, and even across a restart it stays closed. 
> According to what I got told the region should be automatically reassigned to a new server. Please confirm that this is what is expected. If not and the above seems right, then please disregard and close issue.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.