hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "dhruba Borthakur" <dhr...@yahoo-inc.com>
Subject RE: Specifying the positions of replicas
Date Tue, 26 Feb 2008 00:08:16 GMT
Maybe using LOG.warn("function name") will get your stuff into the
namenode log file.

Thanks,
Dhruba


-----Original Message-----
From: Zhu Huijun [mailto:nautilushj@gmail.com] 
Sent: Monday, February 25, 2008 2:32 PM
To: core-dev@hadoop.apache.org
Subject: Specifying the positions of replicas

Hi, all,

Currently our group want to try some new data layout schemes on Hadoop,
but
I encounter some problems.

In Hadoop, each block has 3 replicas by default. It is said in the
documents
that two of them are put randomly in the same rack, and the third is put
at
another rack. This procedure is also clear in the source code of the
class
org.apache.hadoop.dfs.ReplicationTargetChooser.

What I did was adding a simple System.out.println("function name")
clause in
each function in ReplicationTargetChooser, and intended to check which
ones
are called to select the target nodes for replicas. However, when I use
"./bin/hadoop dfs -copyFromLocal" command to copy some files from the
local
file system to HDFS, it seems none of the functions in the
ReplicationTargetChooser is called, because nothing is printed out on
the
screen or the log files.

I preciate if someone can tell me where I can modify to control the
target
nodes of replicas. Thank you!

Have a nice day!

Huijun Zhu

Mime
View raw message