Hi Lucas,
Sounds strange, it should work. As long as you have all the names of
the slave machines in the slaves file the start-mapred script should
ssh into each one and start a tasktracker. Immediate thoughts are: Is
your ssh working? Is hadoop installed on the same directory structure
as on the machine you run the script from (ignore if running from NFS
mount). Have you looked in the error logs on the slave machines, there
might be some useful information there if the script tried to start a
tasktracker and failed? What does the console print when you run
hadoop-mapred.sh? It should say something like "machinex: starting
tasktracker, logging to /somefile/" for each machine listed in your
slaves file.
Good luck,
Ollie
Quoting Lucas Nazário dos Santos <[EMAIL PROTECTED]>:
Hello,
I was reading Hadoop's getting started (
http://wiki.apache.org/lucene-hadoop/GettingStartedWithHadoop), and in the
section named "Starting up a larger cluster" I had a doubt about starting
tasktrackers on datanodes. The tutorial says that after running the
start-dfs.sh script on the node I want as the namenode, I need to execute
the script start-mapred.sh, and this will "bring up the Map/Reduce cluster
with Jobtracker running on the machine you ran the command on and
Tasktrackers running on machines listed in the slaves file".
It actually brings up the Jobtracker on the machine I ran the command on,
but not the tasktrackers on slave machines. I need to enter slave by slave
starting the tasktracker manually, issuing the command "hadoop tasktracker"
what, according to Hadoop's getting started documentation, is something
unnecessary.
Did I misunderstand something?
Thanks in advance,
Lucas
