Hi Henning, On Fri, Sep 7, 2012 at 2:31 PM, Henning Blohm <[email protected]> wrote: > Hi, > > I am using CDH4.01 from a tarball and so far everything works kind of ok > (albeit quite different from CDH3). It seems though that I fail to get an > embedded job history server running when using YARN. On the other hand, > there is no $HADOOP_HOME/sbin/start-historyserver.sh or the like. I can > start the history server using mr-jobhistory-deamon.sh with appropriate > params.
Please use HADOOP_PREFIX instead of HADOOP_HOME. The latter's been deprecated. As you've found out, using the following is the correct way for starting the MR2 Job History Server when starting via a tarball: $ sbin/start-historyserver.sh start historyserver > In other words: Is there a way to have an embedded job history server (small > setup) when using YARN? By "embedded" do you mean a non-separate daemon for MR jobs? Or do you mean start-yarn.sh itself starting the YARN and MR2 (which uses YARN) are two separate set of components, and hence start-yarn.sh doesn't also start the MR-specific job history server. The JHS doesn't apply/work for any other apps of YARN other than MR2 right now, so its not unified. > Thanks, > Henning > > -- Harsh J
