Assuming you are using toolrunner, your conf param should come before your application specific params.
Sent from my iPad On Aug 15, 2010, at 12:55 PM, Jake Vang <[email protected]> wrote: > i seem to be having problems submitting jobs from hadoop using cygwin in > windows (windows 7) to a hadoop multi-node cluster (ubuntu). > > in windows/cygwin, i have created a user called hadoop. this user, hadoop, is > also the user that is created on the master/slave nodes as well. furthermore, > i have set dfs.permissions to false. > > i have an xml file called, hadoop-cluster.xml with the following name-value > pairs. > * fs.default.name=hdfs://hadoop-0:54310 > * mapred.job.tracker=hadoop-0:54311 > > when i run hadoop (using cygwin) i try a command like this: > bin/hadoop my-lib.jar /user/hadoop/input-dir /user/hadoop/output-dir -conf > conf/hadoop-cluster.xml > > but it always fails. the command never passes in the input and output > arguments to the program and it exits. why is this? when i upload this jar > file to a data node (more precisely, the master node, name node), and run > this command then it works (of course, then i don't need to pass in the -conf > param). > > any help is appreciated.
