Hi all,

  I have a Backup Node (BN) and a active NameNode (NN), but if NN failed,
how to recover the NN? Is possible BN replace NN directly? Any resources?

Shen

I think I got it!

Step by step as following:

 1. put two files into hdfs.
 2. kill -9 ${namenode_pid}.
 3. delete the whole ${hadoop.tmp.dir}/dfs directory of NN.
 4. copy the dfs directory of BN to NN.
 5. start-up NN.

It works. Is it correct?

Shen
On Fri, Sep 17, 2010 at 2:58 PM, ChingShen <[EMAIL PROTECTED]> wrote:

> Hi all,
>
>   I have a Backup Node (BN) and a active NameNode (NN), but if NN failed,
> how to recover the NN? Is possible BN replace NN directly? Any resources?
>
> Shen
>

On Sep 17, 2010, at 1:36 AM, ChingShen wrote:

> I think I got it!
>
> Step by step as following:
>
> 1. put two files into hdfs.
> 2. kill -9 ${namenode_pid}.
> 3. delete the whole ${hadoop.tmp.dir}/dfs directory of NN.
> 4. copy the dfs directory of BN to NN.
> 5. start-up NN.
>
> It works. Is it correct?
In my continuing streak of writing content for the wiki, I just filled in a page we've apparently had empty for a while now:

http://wiki.apache.org/hadoop/NameNodeFailover

Note: I'm not fully awake.  It is likely missing stuff.

You may want to check importcheckpoint option too
http://hadoop.apache.org/hdfs/docs/current/hdfs_user_guide.html#Import+Checkpoint

Boris.

On 9/17/10 1:36 AM, "ChingShen" <[EMAIL PROTECTED]> wrote:

I think I got it!

Step by step as following:

 1. put two files into hdfs.
 2. kill -9 ${namenode_pid}.
 3. delete the whole ${hadoop.tmp.dir}/dfs directory of NN.
 4. copy the dfs directory of BN to NN.
 5. start-up NN.

It works. Is it correct?

Shen
On Fri, Sep 17, 2010 at 2:58 PM, ChingShen <[EMAIL PROTECTED]> wrote:

> Hi all,
>
>   I have a Backup Node (BN) and a active NameNode (NN), but if NN failed,
> how to recover the NN? Is possible BN replace NN directly? Any resources?
>
> Shen
>