#puppet-zookeeper

A puppet receipt for Apache Zookeeper. ZooKeeper is a high-performance coordination service for maintaining configuration information, naming, providing distributed synchronization, and providing group services.

Requirements

• Puppet >= 2.7, Puppet 3.x, Puppet 4.x
• Ruby 1.9.3, 2.0.0, 2.1.x
• binary package of ZooKeeper

Basic Usage:

class { 'zookeeper': }

Cluster setup

When running ZooKeeper in the distributed mode each node must have unique ID (1-255). The easiest way how to setup multiple ZooKeepers, is by using Hiera.

hiera/host/zk1.example.com.yaml:

zookeeper::id: '1'

hiera/host/zk2.example.com.yaml:

zookeeper::id: '2'

hiera/host/zk3.example.com.yaml:

zookeeper::id: '3'

A ZooKeeper quorum should consist of odd number of nodes (usually 3 or 5). For defining a quorum it is enough to list all IP addresses of all its members.

class { 'zookeeper':
  servers => ['192.168.1.1', '192.168.1.2', '192.168.1.3']
}

Currently, first ZooKeeper in the array above, will be assigned ID = 1. This would produce following configuration:

server.1=192.168.1.1:2888:3888
server.2=192.168.1.2:2888:3888
server.3=192.168.1.3:2888:3888

where first port is election_port and second one leader_port. Both ports could be customized for each ZooKeeper instance.

class { 'zookeeper':
  election_port => 2889,
  leader_port   => 3889,
  servers       => ['192.168.1.1', '192.168.1.2', '192.168.1.3']
}

Observers

Observers were introduced in ZooKeeper 3.3.0. To enable this feature simply state which of ZooKeeper servers are observing:

class { 'zookeeper':
  servers   => ['192.168.1.1', '192.168.1.2', '192.168.1.3', '192.168.1.4', '192.168.1.5'],
  observers => ['192.168.1.4', '192.168.1.5']
}

Note: Currently observer server needs to be listed between standard servers (this behavior might change in feature).

Set binding interface

By default ZooKeeper should bind to all interfaces. When you specify client_ip only single interface will be used. If $::ipaddress is not your public IP (e.g. you are using Docker) make sure to setup correct IP:

class { 'zookeeper':
  client_ip => $::ipaddress_eth0
}

or in Hiera:

zookeeper::client_ip: "%{::ipaddress_eth0}"

This is a workaround for a a Facter issue.

ZooKeeper service

Use service_provider to override Puppet detection for starting service.

class { 'zookeeper':
  service_provider    => 'init',
  manage_service_file => false,
}

Some reasonable values are:

• init - RHEL6, Debian 7
• upstart - Ubuntu
• systemd - RHEL 7, Debian 8
• runit
• none - service won't be installed

Parameter manage_service_file controls whether service definition should be managed by Puppet (default: false). Currently supported for systemd and init.

Systemd Unit 'After' and 'Want' control

By default the module will create the following Unit section in /etc/systemd/system/multi-user.target.wants/zookeeper.service

[Unit]
Description=Apache ZooKeeper
After=network.target

Both After and Want (omitted when using the module defaults) can be controled using this module.

E.g on CentOS 7 those might have to be configured for 'netwrok-online.target' using the following syntax:

class { 'zookeeper':
   systemd_unit_after => 'network-online.target',
   systemd_unit_want => 'network-online.target',
}

Which will modify the Unit section to look like:

[Unit]
Description=Apache ZooKeeper
Want=network-online.target
After=network-online.target

Parameters

• id - cluster-unique zookeeper's instance id (1-255)
• datastore
• datalogstore - specifying this configures the dataLogDir ZooKeeper config values and allows for transaction logs to be stored in a different location, improving IO performance
• log_dir
• purge_interval - automatically will delete ZooKeeper logs (available since ZooKeeper 3.4.0)
• snap_retain_count - number of snapshots that will be kept after purging (since ZooKeeper 3.4.0)
• min_session_timeout - the minimum session timeout in milliseconds that the server will allow the client to negotiate. Defaults to 2 times the tickTime (since ZooKeeper 3.3.0)
• max_session_timeout - the maximum session timeout in milliseconds that the server will allow the client to negotiate. Defaults to 20 times the tickTime (since ZooKeeper 3.3.0)
• manage_service (default: true) whether Puppet should ensure running service
• manage_service_file when enabled on RHEL 7.0 a systemd config will be managed
• ensure_account controls whether zookeeper user and group will be ensured (set to false to disable this feature)

and many others, see the init.pp file for more details.

If your distribution has multiple packages for ZooKeeper, you can provide all package names as an array.

class { 'zookeeper':
  packages => ['zookeeper', 'zookeeper-java']
}

Hiera Support

All parameters could be defined in hiera files, e.g. common.yaml, Debian.yaml or zookeeper.yaml:

zookeeper::id: 1
zookeeper::client_port: 2181
zookeeper::datastore: '/var/lib/zookeeper'
zookeeper::datalogstore: '/disk2/zookeeper'

Cloudera package

In Cloudera distribution ZooKeeper package does not provide init scripts (same as in Debian). Package containing init scripts is called zookeeper-server and the service as well. Moreover there's initialization script which should be called after installation. So, the configuration might look like this:

class { 'zookeeper':
  packages             => ['zookeeper', 'zookeeper-server'],
  service_name         => 'zookeeper-server',
  initialize_datastore => true
}

Managing repository

For RedHat family currently we support also managing a cloudera yum repo versions 4, and 5. It can be enabled with repo parameter:

class { 'zookeeper':
  repo   => 'cloudera',
  cdhver => '5',
}

Custom RPM repository

Optionally you can specify a custom repository, using a hash configuration.

class { 'zookeeper':
  cdhver     => '5',
  repo       =>  {
    name  => 'myrepo',
    url   => 'http://cusom.url',
    descr => 'description'
  }
}

Java installation

Default: false

By changing these two parameters you can ensure, that given Java package will be installed before ZooKeeper packages.

class { 'zookeeper':
  install_java => true,
  java_package => 'openjdk-7-jre-headless'
}

Install

Librarian (recommended)

For puppet-librarian just add to Puppetfile

from Forge:

mod 'deric-zookeeper'

latest (development) version from GitHub

mod 'deric-zookeeper', git: 'git://github.com/deric/puppet-zookeeper.git'

submodules

If you are versioning your puppet conf with git just add it as submodule, from your repository root:

git submodule add git://github.com/deric/puppet-zookeeper.git modules/zookeeper

Dependencies

• stdlib > 2.3.3 - function ensure_resources is required
• datacat - experimental (might be removed in future releases)

Supported platforms

• Debian/Ubuntu
  • Debian 6 Squeeze: you can get ZooKeeper package from Wheezy or Sid repo.
  • Debian 7 Wheezy: available in apt repository
• RedHat/CentOS/Fedora

Tested on:

• Debian 6 - Squeeze, 7 - Wheezy, 8 - Jessie
• Ubuntu 12.04.03 LTS, 14.04
• RHEL 6, RHEL 7, CentOS 6