On Mon, Nov 16, 2009 at 9:57 AM, Jeff Zhang <zjf...@gmail.com> wrote: > look at the logs of job tracker, maybe you will get some clues. > > > Jeff Zhang > > > > On Mon, Nov 16, 2009 at 6:45 AM, Prabhu Hari Dhanapal < > dragonzsn...@gmail.com> wrote: > >> Hi all, >> >> I just installed Hadoop(single node cluster) and tried to start and stop >> the >> nodes , and it said >> no jobtracker to stop , no namenode to stop >> >> however the tutorial i used suggest that jobtracker and namenodes should >> also have started ? Why does this happen? >> am i missing something? >> >> >> http://www.michael-noll.com/wiki/Running_Hadoop_On_Ubuntu_Linux_(Single-Node_Cluster)<http://www.michael-noll.com/wiki/Running_Hadoop_On_Ubuntu_Linux_%28Single-Node_Cluster%29> >> >> >> had...@pdhanapa-laptop:/home/pdhanapa/Desktop/hadoop/bin$ jps >> 20671 Jps >> 20368 DataNode >> 20463 SecondaryNameNode >> >> had...@pdhanapa-laptop:/home/pdhanapa/Desktop/hadoop/bin$ ./stop-all.sh >> no jobtracker to stop >> localhost: no tasktracker to stop >> no namenode to stop >> localhost: stopping datanode >> localhost: stopping secondarynamenode >> >> >> >> >> -- >> Hari >> >
The issue here is that these resources failed to start. What happens here is as soon as the java process is started the system returns an ok status to the script. However the processes die moments later as they start up. For example if you start the namenode, script returns ok, namenode runs and realizes its dfs.name directory is not formatted. Then it stops. Generally after starting a hadoop process, tail the log it creates for a few seconds and make sure it REALLY starts up. Really the scripts should do more pre-startup checking, but the scripts could not test for every possible condition that could cause hadoop not to start. Also for long running deamons the pid files are written to /tmp see bin/hadoop-daemon.sh If something is cleaning /tmp stop arguments are unable to find the pid. That is shell scripting for you :) Edward
