Getting started with HBase

Web 2.0, AJAX, HTML5, and now Big Data. Buzzwords, buzzwords everywhere. However, Big Data means data, and these large amounts of data need to be stored somewhere. Apache HBase is a distributed data store well suited for the task. As I've been playing around with Vagrant and Ansible recently, I've created a virtual machine environment for quick-and-dirty HBase setup. See vagrant-hbase at my GitHub.


Note: this blog entry is more or less the same as README contents. Lazy writer, am I?

This Vagrant project sets up a Ubuntu 12.04 (precise) virtual machine and provisions it for HBase development. Ansible is used for VM provisioning.

After running vagrant up, a single HBase node is set up in standalone mode (using local file system, not HDFS). A Thrift server is also running, allowing access from languages outside of the JVM.

Getting started

  1. install Ansible on your host machine
  2. git clone && cd vagrant-hbase && vagrant up
  3. ????
  4. PROFIT!

Note: Ansible < 1.2 may not work due to some bugs in parsing Jinja2 syntax in playbooks. Ubuntu 13.04 has only Ansible 1.1 in repositories, therefore it's best to use sudo pip install ansible to get a more recent version.

Network and ports

The guest machine has a private IP address HBase-related ports are forwarded according to the following rule:

hostPortNumber = guestPortNumber + 100

For example, HBase web UIs are available from the host machine at (Master) and (RegionServer).

Initial data

The create_test_table.rb file can be loaded into HBase shell to define a dead simple test table with one column family cf. The table looks like this:

| rowkey |  cf:col1  |
|  row1  |   value1  |

To load the data, execute:

vagrant ssh
cd hbase-0.94.9/
./bin/hbase shell /vagrant/data/create_test_table.rb


A Jython interpreter is also installed in the VM. To access Java HBase API from Jython, you need to set the CLASSPATH environment variable correctly. HBase CLI tool can help in that matter.

vagrant ssh
cd hbase-0.94.9/
export CLASSPATH=`./bin/hbase classpath`

This should drop you into Jython REPL where you can import Java classes and use them in a more Pythonic way. An example script that reads the data loaded with create_test_table.rb is at /vagrant/

Python and HappyBase

An alternative approach to access HBase from Python is to use a regular (CPython) interpreter and Thrift bindings. A Python library called HappyBase hides away all the necessary plumbing behind a simple, clean interface to your data. For your convenience, this package is installed into system-wide Python libraries when provisioning the guest machine.

An example Python session on the guest machine may look like this (assuming initial data were loaded):

vagrant@vagrant-hbase:~$ python
Python 2.7.3 (default, Apr 10 2013, 05:46:21)
[GCC 4.6.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import happybase
>>> conn = happybase.Connection()
>>> print conn.tables()
>>> table = conn.table('test')
>>> print table.row('row1')
{'cf:col1': 'value1'}