Mailing-List: contact issues-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Message-ID: <6588289.58791276317912943.JavaMail.jira@thor>
Date: Sat, 12 Jun 2010 00:45:12 -0400 (EDT)
From: "stack (JIRA)" <jira@apache.org>
To: issues@hbase.apache.org
Subject: [jira] Created: (HBASE-2717) Bring up master UI even if .META. and
 -ROOT- not deployed; add status message to front of master ui (recovering
 server X, etc.)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

Bring up master UI even if .META. and -ROOT- not deployed; add status message to front of master ui (recovering server X, etc.)
-------------------------------------------------------------------------------------------------------------------------------

                 Key: HBASE-2717
                 URL: https://issues.apache.org/jira/browse/HBASE-2717
             Project: HBase
          Issue Type: Bug
            Reporter: stack


I went to visit some fellas today who'd yanked the hdfs out from under their 80-node hbase because of an emergency.  Interestingly, the maser was reporting 80 dead servers and no live servers.  It couldn't recover because hdfs was gone.

On restart, the master had 80 server's logs to replay.  There is a code-path we follow on startup where all is suspended until ALL logs are split which in this case of 80*(15-20 wals each) was taking a long time.  To the administrator who is just getting started on hbase, tailing master log, it didn't look like it was making progress so he kept killing master, and so on.

The gentleman suggested that if the web UI had come up and said roughly what it was about w/ perhaps a bit of a progress message he'd not have been so disorientated.

This issue is about making it so web ui comes up even if no regionservers -- i.e. meta and root not deployed -- and adding a rough status message (splitting logs on dead server X, etc.)

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