Hello, Not sure if this scope is too narrow for the whirr ml or not. I've come across the whirr-cm <https://github.com/cloudera/whirr-cm>library and I'm exploring whether I can extend it, rather than implement a client using the raw CM Java API, given the breadth features available in whirr and it's ease of use.
What I'm wondering is if it's possible to subclass ClusterSpec to be used with my deployment. What I'd like to do is create a whirr properties file like this one <https://raw.github.com/cloudera/whirr-cm/master/cm-ec2.properties>, but stripped of any provider/identity info and instead supply the host/port/username/password of my CM server, plus the hostnames/IPs and desired services/roles configuration. In other words, I'm hoping to leverage whirr to pass a Hadoop Cluster topology to an installed, but not yet initialized, Cloudera Manager Server. Unlike the Whirr cloud EC2/Rackspace paradigms, these machines are already provisioned, and their hostnames/IPs known. Is this possible with whirr? I'd imagine that if it is, I'd have to subclass one or more of the whirr classes, but I'm not sure where to begin. If anyone has any advice (even if it's, 'don't go there'), I would appreciate it. -Joe
