The available playbooks can be used to:
-
setup a CDH based Hadoop cluster via Cloudera Manager and integrated Cloudera Search
-
[planned]
- setup a CDH based Hadoop cluster without the usage of Cloudera Manager
- install Cloudera Manager on top of an already running Hadoop cluster
Cloudera: www.cloudera.com Ansible: www.ansibleworks.com
Prerequisites
- oracle jdk1.6 is highly recommended by Cloudera, ensure you have no other java version on your cluster nodes. The playbook will install the corresponding oracle jdk
- the playbook will connect as user root to the cluster nodes. If you have a desired user, edit the header section of the playbooks accordingly
- all your nodes are installed with a plain Debian Squeeze. Cloudera's packages seem to work on Wheezy also, but I have not tested it currently.
- ansible is installed on your client, e.g. your workstation, as described under http://www.ansibleworks.com/docs/gettingstarted.html, and this Github repo has been checked out
What to adjust for your environment ?
-
edit config/hosts and replace the node names hadoop-pg-[1-5] by the DNS names of your nodes. Replace the names for the roles 'namenode', 'jobtracker', 'clouderamanager' accordingly
-
edit vars/base.yml and set the corresponding hostname of the node which acts as Cloudera Manager node
Ansible usage hints
- add parameter -v/-vv/-vvv for increasing verbose output of the commands ansible is executing
- remove parameter -k if you have passwordless ssh connection from your client to the cluster nodes
How to handle the playbooks ?
- setup a CDH based Hadoop cluster via Cloudera Manager including integrated Cloudera Search
- a detailed step-by-step instruction is available at http://www.ymc.ch/en/how-to-install-cloudera-manager-and-cloudera-search-with-support-from-ansible