Greetings,
Apache Ambari 1.5.1 (latest release) is available in F20 and rawhide. It is a Hadoop cluster management project with a centralized server, web console and agent framework. The project originated from Hortonworks and to-date features stack definitions and bits from their repos, the latest being HDP 2.1. Supported agent OS include the RHEL/CentOS family (6 and 5). These agents can be installed manually and self-register, or be provisioned to an "empty" node by the server via scp and bootstrap Python scripts.
Interoperability between the Fedora Ambari server and Hortonworks Ambari agents from RHEL6 and CentOS6 appear to work correctly for the manual registration use case. The server provisioning use case needs some more work and likely patches upstream due to some lingering inflexibility with their software requirements. Currently, they enforce a strict version alignment of the agents to server sprinkled in various pockets of their configuration and Python code base.
The Fedora agent is packaged but has unresolved interop issues when deploying even a basic stack of HDFS and Zookeeper. Beyond these agent issues, there will be other technical considerations which may lead me to attempt to define and implement a Fedora stack obviously based on our expanding set of Big Data packaged components. Stay tuned but in the meantime feel free to tinker if you like (and raise BZ) as I iterate on the 1.5.x package some more.
http://koji.fedoraproject.org/koji/packageinfo?packageID=18392
\Pete
bigdata@lists.fedoraproject.org