GitLab

The mapreduce terasort test was filling up /tmp so I tried setting the temporary directory to /mnt to give HDFS more temp space at the cost of space for HDFS data

Needed for previous change to logging

This allows one to use compound targeting only for spark

this is similar to hdfs, mapred, and yarn

to see if that stops the /tmp/hadoop-yarn dir from being created. It is not being set with the correct permissions to allow general users to run jobs and I want to know where the directories are being created.

I figured out how to set this via grains. It will now be set using cloudInit to edit a grain file and set the location of the hdfs file system

Added https://github.com/cgeroux/hadoop-formula.git to https://github.com/cgeroux/hadoop-formula.git

enable source_url from fileserver

#36: update source URLs, set default version to apache-2.7.1

Make HA namenode init more robust in real distributed environments

#17 add support for HDFS namenode HA

#28 safer PID directory handling in init script

#31: fix hdp2.2 download URL customization

Fix nodemanager target

Strip unicode chars from example

nodemanager_target incorrectly fetched grain resourcemanager_target

The pillar example contained non-ascii characters that
if copy/pasted into an SLS file cause strange failures