Hi,
We have a problem on our hadoop 2.2.0 cluster. We're running mapreduce2
applications.
The problem is the following: When "mapreduce.job.classloader" property
is set to true, resource manager links to application master don't work.
These links are of the form
"http://<RM_host>:<RM:port>/proxy/application_<app_id>". Normally they
take me to a web page listing all the jobs managed by the application
master (1 for mapreduce jobs) with links to job pages of the running
job. But with above property set to true, these links send me to a
default web page listing an empty directory.
It seems that the resource manager/application master can't find needed
files in given directory do construct and display a proper web page.
Did anyone have this problem? Do you know if and how this can be fixed?
