Description
I think there's a typo in the hadoop-config.sh which broke slave start in cluster mode.
To reproduce it :
/usr/local/hadoop/sbin/hadoop-daemons.sh --config /etc/hadoop/conf --hosts slaves start datanode
here's the patch :
— hadoop-dist/target/hadoop-2.4.0/libexec/hadoop-config.sh 2014-04-22 12:20:45.000000000 -0400
+++ hadoop-dist/target/hadoop-2.4.0/libexec/hadoop-config.sh 2014-04-22 12:21:00.000000000 -0400
@@ -93,7 +93,7 @@
if [ "--hosts" = "$1" ]
then
shift
- export HADOOP_SLAVES="${HADOOP_CONF_DIR}/$$1"
+ export HADOOP_SLAVES="${HADOOP_CONF_DIR}/$1"
shift
elif [ "--hostnames" = "$1" ]
then