Hi  team,

The current trunk and the upcoming 0.7.2 release has support for Apache
Hadoop, Apache HBase, Apache ZooKeeper, YARN, CDH3 and CDH4. Unfortunately
it's not obvious how a recipe should look like for each option because
we've started to abuse the ability to override install / configure
functions.

IMHO the ability to override a function should be considered an advanced
feature - not a common mechanism for selecting a service flavor.

I suggest we take some time to make the source code clean for 0.8.0. I
think we should have a set of distinct roles for each flavor:

zookeeper, cdh3-zookeeper, cdh4-zookeeper etc.
cdh3-namenode, cdh4-namenode (yarn) etc.

What do you think?

-- Andrei Savu

Reply via email to