HI All,
I'm a little baffled by the new web UI in Hadoop 2.0. In particular, I
don't see an obvious way of viewing the task specific counters or the
task status's that my jobs are setting. Previously, this was all
presented cleanly in the TaskTracker UI but now the
ApplicationManager/HistoryServer obscure this information.
I've found that
http://host:8088/proxy/<app_id>/mapreduce/jobcounters/<job_id> presents
the counters over *all* tasks in the job and similarly
http://host:8088/proxy/<app_id>/mapreduce/attempts/<job_id>/m/RUNNING
presents the current state of the running tasks. However neither of
these link to task specific counters, and in the running task display,
the "STATE" column seems completely unassociated with the task's status.
Is there an easy way to view all this information without writing my own
webapp around the ApplicationManager/HistoryServer?
I'm also not sure how to view these two web ui's via alternate
hostnames, as was done automatically in the older versions of Hadoop.
In my hdfs and yarn configurations, i set the hostname as krackenmaster,
which uses one network interface. At the same time, i'd like to view
these ui's using a second, and more public, hostname such as kracken.com
that just points to the master node in my cluster. Would setting up the
Web Application Proxy (described in
http://hadoop.apache.org/common/docs/r0.23.1/hadoop-yarn/hadoop-yarn-site/WebApplicationProxy.html)
accomplish this? As a note, I can view the NameNode UI from kracken.com
without any problem.
Thanks!
--Keith