hadoop-common-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gautam Kowshik <gaut...@yahoo-inc.com>
Subject Re: Data Node failover
Date Wed, 10 Jan 2007 09:52:35 GMT
We have replication in place to account for cases when a datanode is not 
reachable. The namenode(master) starts replicating files that were on 
that data node. You can also tell the dfs to maintain more copies(than 
usual) of certain important files using a replication factor. Read more 
bout it here :
http://lucene.apache.org/hadoop/hdfs_design.html#Data+Replication

What kinda/how much data are u using on your 3 node dfs. The namenode 
can replicate 50 blocks per second on average, so i don't think time is 
a problem. A problem could be that the dfs is not able to maintain enuf 
replicas. Could u mention what exception u got?
-Gautam.

Sarvesh Singh wrote:
> Hi,
>
> I have a hadoop cluster of 3 instances, when I kill data node process 
> on one
> of the slave machine,
> failover does not seem to work. Another slave machine does the copy of 
> DFS
> block for 7-10 minutes
> but client program bombs with an exception after 7-10 minutes.
> Do we have data node failover implemented in Hadoop??
>
> Thanks
> Servesh
>


Mime
View raw message