This allows for creating an installer for Hadoop for Mac OS X that attempts to integrate Hadoop in a typical manner with the OS:
- Most files are put in
/Library/Cloudera
. launchd
will be used for stopping and starting the various daemons.- Paths are added to
PATH
by way of/etc/paths.d
.
- Daemons don't correctly start up for Hadoop and I haven't done anything yet for ZooKeeper.
- I haven't put HBase, Hive, Flume or Hue in here yet.
- Permissions are not quite right yet for various files. This is pretty frustrating.
- It would be nice to automate setting
JAVA_HOME
for those things that need it. (This is just a matter of settingJAVA_HOME
to the output of/usr/libexec/java_home
.) - I haven't done anything to deal with updating to newer versions yet.
- To build the installer, you download the various
.tar.gz
files from Cloudera, unpack them and put them in the right places underdist
such asdist/Hadoop/Library/Cloudera/hadoop-0.20.2+737
. - There are symlinks without version numbers (and capitalized) that all go into
/Library/Cloudera
on the target system. If the version being installed changes, then the symlink must also be updated. - I will have another script added at some point that handles setting up the permissions correctly once I figure out what all is needed there.
- I run the
CDH3.pmdoc/*.xml
files throughxmllint
to format them nicer for diffs with theprettyxml.sh
script.