Skip to content

npoggi/ansible-hadoop

 
 

Repository files navigation

ansible-hadoop

These Ansible playbooks will build a Hadoop cluster.

You can pre-build a Rackspace cloud environment or run the playbooks against an existing environment.


[Installation] (id:installation)

See INSTALL-ENV.md for installation and build instructions.

[Requirements] (id:requirements)

  • Ansible >= 2.0.

  • Expects RHEL/CentOS 6/7 or Ubuntu 14 hosts.

  • Building the Rackspace Cloud environment requires the pyrax Python module: pyrax link.

[Features] (id:features)

  • It installs Hortonworks Data Platform using Ambari Blueprints.

  • It supports static inventory if the environment is pre-built (inventory/static file).

  • The data drives can be customized and can be put on top of Cloud Block Storage (partitioning is automatic).

  • If there are 2 or 3 masternodes, it will also enable HA NameNode.

  • Memory settings are scaled with the hardware configuration of the nodes.

[Inventory] (id:inventory)

  • The cloud environment requires the standard pyrax credentials file that looks like this:

    [rackspace_cloud]
    username = my_username
    api_key = 01234567890abcdef
    

    This file will be referenced in playbooks/group_vars/all (the rax_credentials_file variable).

    By default, the file is expected to be: ~/.raxpub.

  • When provisioning HDP on existing infrastructure edit inventory/static and add the nodes.

[Configuration files] (id:configuration)

To customize, change the variables under playbooks/group_vars folder:

  1. playbooks/group_vars/all: contains global cluster and cloud settings
  2. playbooks/group_vars/master-nodes: master-nodes configuration
  3. playbooks/group_vars/slave-nodes: slave-nodes configuration
  4. playbooks/group_vars/edge-nodes: edge-nodes configuration

For a one-node cluster, set cloud_nodes_count in master-nodes to 1 and cloud_nodes_count in slave-nodes to 0.

[Scripts] (id:scripts)

###provision_rax.sh

To provision a cloud environment, run the provision_rax.sh script after you've customized the variables under playbooks/group_vars:

bash provision_rax.sh

###bootstrap* and hortonworks*

Similarly, run the bootstrap and hortonworks scripts (in this order), depending what type of environment you have.

  • For a Rackspace Cloud environment:

    bash bootstrap_rax.sh
    bash hortonworks_rax.sh
    
  • For static / prebuilt environments:

    bash bootstrap_static.sh
    bash hortonworks_static.sh
    

[Accessing Ambari] (id:ambari)

Once you are at this point you can see progress by accessing the Ambari interface (the ambari-node will be the last host that ran a play).

The provided Ansible playbook will only open the firewall if you've added your workstation IP to allowed_external_ips variable in the playbooks/group_vars/all file.

Alternatively, you can access Ambari by either opening the firewall manually or by opening a socks proxy with the following command:

ssh -D 12345 root@ambari-node

You will need to modify your browser settings to use socks proxy localhost and port 12345.

You'll then be able to navigate to http://ambari-node:8080 in your configured browser and access all subsidiary links.

[Ansible-Hadoop History] (id:history)

As with many projects this code is the end result of a lot of effort from individuals not properly represented by a simple commit history.

Rackspace started deploying Hadoop on dedicated gear for customers more than a year ago in a very manual process. This process landed with myself and these Rockstars:

Joe Engel (Racker Emeritus)

Mark Lessel

Alexandru Anghel

All of whom wrote a lot of the automation for deploying Hadoop on customer gear at Rackspace.

Today with a pile of customers under our belt and many more all the time, we wanted to share our efforts with the world by publishing this code which you can also use to deploy Hadoop in various ways at Rackspace.

This of course is only the beginning!

I hope this project evolves and inspires even more Rockstars to find ways to contribute.

About

Ansible playbooks for deploying Hortonworks Data Platform

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 74.2%
  • Shell 25.8%