Bento Cluster Docker

A prototype for a Bento cluster (self contained HDFS/YARN/HBase/Cassandra environment). Runs CDH 5 and Cassandra 2 in pseudo-distributed mode.

Installation

The Bento Cluster can be installed using pip:

pip3 install kiji-bento-cluster

To upgrade your installation:

pip3 install --upgrade kiji-bento-cluster

Usage

The Bento Cluster is controlled from the bento script. bento help will show a listing of all available bento options.

In this section strings in code blocks situated in between angle braces (<, >) represent fields to be filled in by the user.

Many of the commands below make usage of the --bento-name= flag. If it is not specified, the name 'bento' is used.

A simple workflow

Pull

Run the bento pull command to fetch the latest bento docker image from dockerhub. Note: this operation may pause for a long period of time without any output.

If a specific platform version is desired, specify it with the --platform-version= flag:

bento pull --platform-version=cdh5.1.2

Create

Run the bento create command to create a new bento instance. A name can be specified with the --bento-name= flag:

bento --bento-name=<my_very_own_bento> create

This command can be validated by running the bento list command with the --all option (displays running bento instances as well as ones that aren't running):

bento list --all

Start

Run the bento start command to start a created bento instance. Data already existing within bento instances being started is preserved. To start a specific bento instance use the --bento-name= flag:

bento --bento-name=<my_very_own_bento> start

This command will block until hdfs has been initialized. The list of running bento instances can be printed by running:

bento list

The status of a particular bento instance can also be printed by running:

bento --bento-name=<my_very_own_bento> status

Bento cluster uses supervisor to bring up and watch its various services. The status of each service can be observed by accessing the supervisor web UI available at: http://<my_very_own_bento>:9001/.

After starting a bento instance hadoop/hbase client configuration files will be written to:

~/.bento/<bento-instance-name>/hadoop/
~/.bento/<bento-instance-name>/hbase/

To use a particular bento instance source its 'bento-env.sh' script in:

~/.bento/<bento-instance-name>/bento-env.sh

Stop

Run the bento stop command to stop running processes of a bento (not delete its state/data). To stop a specific bento instance use the --bento-name= flag:

bento --bento-name=<my_very_own_bento> stop

This command will wait until the bento container has stopped.

Rm

Run the bento rm command to delete a bento instance. To delete a specific bento instance use the --bento-name= flag:

bento --bento-name=<my_very_own_bento> rm

Use of Sudo

Bento may ask for sudo permissions when creating or starting a container by default. This is necessary to update /etc/hosts with an entry for the container. This can be handled in several different ways:

• Skip adding a dns entry for the bento instance by using the --skip-hosts-edit (-e) flag with bento create or bento start.

• Add a sudoers rule allowing members of the bento group to run the update-etc-hosts script as root without providing a sudo password:

  bento setup-sudoers
  

  This will also copy the update-etc-hosts script to /usr/local/bin/. Note for Mac users: the /etc/sudoers.d directory is not included by default but can be enabled by adding the #includedir /etc/sudoers.d directive in /etc/sudoers.

Note that HBase and some Hadoop functionality will not work if the container's hostname can not be resolved.

Maven

See bento-maven-plugin.

Requirements

Linux Host

Requires:

1. Docker. Docker requires Linux kernel 3.8 or above.

OS X Host

Requires:

1. boot2docker

2. A network route to the bento box:

       sudo route add $(bento ip)/16 $(boot2docker ip 2> /dev/null)

   If you are using OS X and planning on running any MapReduce jobs, you will likely want to allocate boot2docker more than its default 2 GB of RAM. You can find directions for doing so on the github page for the boot2docker CLI. In short, do the following:

   • Create a boot2docker profile with the default settings:

       boot2docker config > ~/.boot2docker/profile

   • Update the line in ~/.boot2docker/profile to increase the amount of memory from 2048 to perhaps
   • Sanity check your new settings by running boot2docker config again.
   • Destroy your old boot2docker VM and start again:

       boot2docker destroy
       boot2docker init
       boot2docker up

   • Validate your new settings:

       boot2docker info