Are you setting HADOOP_CONF_DIR in hadoop-env.sh? Do you have YARN_HOME set in hadoop-env.sh? Is your yarn-env.sh just the standard one from ./hadoop-mapreduce-project/hadoop-yarn/conf/yarn-env.sh?
On 1/9/12 6:16 AM, "Praveen Sripati" <[EMAIL PROTECTED]> wrote:
I am trying to setup 0.23 on a cluster and am stuck with errors while starting the NodeManager. The slaves file is proper and I am able to do a password-less ssh from the master to the slaves. The ResourceManager also starts properly.
On running the below command from the master node.
>> bin/yarn-daemons.sh start nodemanager
I get the following error from the yarn script
>> praveen-laptop: No HADOOP_CONF_DIR set.
>> praveen-laptop: Please specify it either in yarn-env.sh or in the environment.
The following is set in .bashrc for all the slaves
>> export HADOOP_DEV_HOME=/home/praveensripati/Installations/hadoop-0.23.0
>> export HADOOP_CONF_DIR=$HADOOP_DEV_HOME/conf
For some reason the HADOOP_CONF_DIR and other variables are not visible in the yarn-daemon.sh (could verify if from the echo statements also) which calls the yarn script.
I am able to run the below command successfully from the terminal, but the yarn-daemons.sh (plural) command exits with the above error.
>> ssh praveen-laptop cd /home/praveensripati/Installations/hadoop-0.23.0/bin/.. ; /home/praveensripati/Installations/hadoop-0.23.0/bin/yarn-daemon.sh --config /home/praveensripati/Installations/hadoop-0.23.0/conf start nodemanager
I looked at the code and could not figure it out. Did I miss anything or is there a bug in the code?