I have an hadoop cluster run by amazon and I have access to the master node in ssh
therefore i set up tunnels to access from localhost all relevant services

the ports open on the amazon master node are

9000 -> fs.default.name
9001 -> mapred.job.tracker
9100 ->
9101 ->

and when in kettle I create an hdfs file input step the connect operation to hdfs:// returns the error in the subject

I verified the hadoop jar and the commons one
What else? where to look for?