Hadoop Mapreduce tasktrackers keep ignoring HADOOP_CLASSPATH. Zookeeper trying to connect to localhost rather than cluster address - hadoop

Question

I have a Hadoop cluster (Cloudera CDH4.2) with 5 datanodes. I'm trying to run a MapReduce job which creates an HBaseConfiguration object. The tasktracker attempts fail, because they're trying to connect to localhost:2181 rather than the address of the actual zookeeper installation.
I'm aware that this is because the tasktrackers are not being supplied with the correct classpath containing the hbase configuration. However, if I run the job like so:
HADOOP_CLASSPATH=`/usr/bin/hbase classpath` hadoop jar myjar.jar
The documentation indicates this should solve the problem. The first entry in hbase classpath is /usr/lib/hbase/conf which is a symlink to /etc/hbase/conf, so in theory, this should add the hbase configuration into the HADOOP_CLASSPATH variable.
However, the logs from the tasktracker show this:
2013-08-14 12:47:24,308 INFO org.apache.zookeeper.ZooKeeper: Client environment:java.class.path=<output of `hadoop classpath`>
....
2013-08-14 12:47:24,309 INFO org.apache.zookeeper.ZooKeeper: Initiating client connection, connectString=localhost:2181 sessionTimeout=180000 watcher=hconnection
.....
2013-08-14 12:47:24,328 WARN org.apache.zookeeper.ClientCnxn: Session 0x0 for server null, unexpected error, closing socket connection and attempting reconnect
java.net.ConnectException: Connection refused
So, for some reason, the tasktrackers are completely ignoring my effort to set HADOOP_CLASSPATH to hbase classpath. The documentation (http://hbase.apache.org/apidocs/org/apache/hadoop/hbase/mapreduce/package-summary.html#classpath) states this should just work. What's wrong?
I'm aware I could work around this by explicitly specifying the zookeeper quorum address in the jar code, but I need this jar to be portable, and pick up the local configuration without recompiling, so I don't see hard coding the address as a viable option.

Answer 1

if you made java programming:
conf.set("hbase.zookeeper.quorum", "server1,server2,server3");
conf.set("hbase.zookeeper.property.clientPort", "2181");
if you used command:add -Dhbase.zookeeper.quorum
sudo hadoop jar /opt/cloudera/parcels/CDH-4.3.0-1.cdh4.3.0.p0.22/lib/hbase/hbase.jar rowcounter -Dhbase.zookeeper.quorum=server1,server2,server3 hly_temp