Philip Zeyliger created BIGTOP-1054:
---------------------------------------
Summary: HADOOP_MAPRED_HOME is cumbersome
Key: BIGTOP-1054
URL: https://issues.apache.org/jira/browse/BIGTOP-1054
Project: Bigtop
Issue Type: Bug
Reporter: Philip Zeyliger
Currently, to run "yarn jar myjob.jar" you have to set the following in
hadoop-env.sh.
{code}
HADOOP_MAPRED_HOME=/usr/lib/hadoop-mapreduce
{code}
This is a bit annoying if you're dealing with multiple Hadoop client
configurations. I'd like to propose the following for consideration:
# If the user is using "yarn" (the binary) we default HADOOP_MAPRED_HOME to the
implementation of MR2 that we shipped with. Since the {{yarn}} binary knows
its name, and knows the relative path to hadoop-mapreduce, this would be an
easy default to implement.
# We have a softer way to trigger "prefer yarn" rather than specifying the
directory. E.g., HADOOP_MAPRED_PREFER_MR2 This makes it easier to switch
betwen the two without remember the details of where the directory is.
Thoughts!
Thanks!
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
