I have found the problem. There has a error global parameter “HADOOP_CONF_DIR” in "/etc/profile" that is configed by others, and this parameter “HADOOP_CONF_DIR” lead to an unexpected path. So I got the error, even I use “hadoop fs -ls” in these new-adding machine, it shows files in the local file system. When I cancelled this parameter, all works normal.
Thanks. 2011/2/17 James Ram <hadoo...@gmail.com> > Hi > > Try formating the new datanode and then run the hadoop-daemon.sh. > > With Regards, > RJ > > > On Wed, Feb 16, 2011 at 3:50 PM, Jameson Li <hovlj...@gmail.com> wrote: > >> Hi, >> >> My new adding datanodes are working well. >> >> But I just use start-dfs.sh in the namenode and hadoop-daemons.sh start >> datanode in the namenode can start the new adding datanode, I can't use >> hadoop-daemon.sh start datanode in the new adding datanode to start itself, >> and when I do that it will return a null pointer exception. >> >> I have read the mail-archives "New Datanode won't start, null pointer >> exception" reported by Scott, and the url is: >> http://www.mail-archive.com/hdfs-user@hadoop.apache.org/msg00271.html >> >> But there has not have a really result. Scott at last just said:"A >> stop-all/*start*-all on the cluster got it to *start*." >> >> Thanks, >> Jameson. >> > >
