hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From <sallonch...@hotmail.com>
Subject Re: Tables Miss after Adding New Slave
Date Thu, 25 Mar 2010 02:58:42 GMT
Hi, Stack. I use the similar way to datanode: 
"${HADOOP_HOME}/bin/hadoop-daemon.sh start tasktracker". It works! Thanks a 
lot!

Xianglong

--------------------------------------------------
From: <sallonchina@hotmail.com>
Sent: Thursday, March 25, 2010 10:53 AM
To: <hbase-user@hadoop.apache.org>
Subject: Re: Tables Miss after Adding New Slave

> Hi, Stack. Last time you told me how to start the new added datanode and 
> regionserver and I followed your suggestions. It works well. But another 
> question, I saw that tasktracker was not started. So how should I run 
> tasktracker in the new added node? Thanks a lot!
>
> Xianglong
>
> --------------------------------------------------
> From: "Stack" <stack@duboce.net>
> Sent: Wednesday, March 10, 2010 11:59 PM
> To: <hbase-user@hadoop.apache.org>
> Subject: Re: Tables Miss after Adding New Slave
>
>> 2010/3/10  <sallonchina@hotmail.com>:
>>> Hi, everyone. I set up a hadoop cluster with 3 nodes: one mater and two 
>>> slaves. Recently I want to add a new node as a slave to it. After 
>>> configuration, the HDFS and HBase work, however, I found some tables in 
>>> HBase previously disappear. I browse hbase directory in HDFS and found 
>>> directories of the missing tables. Does anyone encounter the same 
>>> situations or is there someone can help me to recover these tables? 
>>> thanks!
>>>
>>
>> Perhaps the node carrying the .META. region -- the catalog table that
>> has table info -- was killed before it had time to flush its edits so
>> added table edits were lost?
>>
>> You could playing with the ./bin/add_table.rb script.  See the head of
>> the script for how to use it.  This should restore your tables by
>> reading the content of the filesystem adding back the missing edits to
>> .META..
>>
>> In future, be careful how you close down your cluster.  You could try
>> flushing .META. first before running shutdown by doing the following
>> in the shell:
>>
>> hbase> flush ".META."
>>
>> St.Ack
>>
> 

Mime
View raw message