Puppet Keepalived

Build status

Requirements

• concat module

Tested on...

• Debian 7 (Wheezy)
• Debian 6 (Squeeze)
• RHEL 6

Example usage

Basic IP-based VRRP failover

This configuration will fail-over when:

a. Master node is unavailable

node /node01/ {
  include keepalived

  keepalived::vrrp::instance { 'VI_50':
    interface         => 'eth1',
    state             => 'MASTER',
    virtual_router_id => '50',
    priority          => '101',
    auth_type         => 'PASS',
    auth_pass         => 'secret',
    virtual_ipaddress => [ '10.0.0.1/29' ],
    track_interface   => ['eth1','tun0'], # optional, monitor these interfaces.
  }
}

node /node02/ {
  include keepalived

  keepalived::vrrp::instance { 'VI_50':
    interface         => 'eth1',
    state             => 'BACKUP',
    virtual_router_id => '50',
    priority          => '100',
    auth_type         => 'PASS',
    auth_pass         => 'secret',
    virtual_ipaddress => [ '10.0.0.1/29' ],
    track_interface   => ['eth1','tun0'], # optional, monitor these interfaces.
  }
}

Add floating routes

node /node01/ {
  include keepalived

  keepalived::vrrp::instance { 'VI_50':
    interface         => 'eth1',
    state             => 'MASTER',
    virtual_router_id => '50',
    priority          => '101',
    auth_type         => 'PASS',
    auth_pass         => 'secret',
    virtual_ipaddress => [ '10.0.0.1/29' ],
    virtual_routes    => [ { to  => '168.168.2.0/24', via => '10.0.0.2' },
                           { to  => '168.168.3.0/24', via => '10.0.0.3' } ]
  }
}

Detect application level failure

This configuration will fail-over when:

a. NGinX daemon is not running
b. Master node is unavailable

node /node01/ {
  include ::keepalived

  keepalived::vrrp::script { 'check_nginx':
    script => '/usr/bin/killall -0 nginx',
  }

  keepalived::vrrp::instance { 'VI_50':
    interface         => 'eth1',
    state             => 'MASTER',
    virtual_router_id => '50',
    priority          => '101',
    auth_type         => 'PASS',
    auth_pass         => 'secret',
    virtual_ipaddress => '10.0.0.1/29',
    track_script      => 'check_nginx',
  }
}

node /node02/ {
  include ::keepalived

  keepalived::vrrp::script { 'check_nginx':
    script => '/usr/bin/killall -0 nginx',
  }

  keepalived::vrrp::instance { 'VI_50':
    interface         => 'eth1',
    state             => 'BACKUP',
    virtual_router_id => '50',
    priority          => '100',
    auth_type         => 'PASS',
    auth_pass         => 'secret',
    virtual_ipaddress => '10.0.0.1/29',
    track_script      => 'check_nginx',
  }
}

Global definitions

class { 'keepalived::global_defs':
  ensure                  => present,
  notification_email      => 'no@spam.tld',
  notification_email_from => 'no@spam.tld',
  smtp_server             => 'localhost',
  smtp_connect_timeout    => '60',
  router_id               => 'your_router_instance_id',
}

Soft-restart the Keepalived daemon

class { '::keepalived':
  service_restart => 'service keepalived reload',     # When using SysV Init
  # service_restart => 'systemctl reload keepalived', # When using SystemD
}

Opt out of having the service managed by the module

class { '::keepalived':
  service_manage => false,
}

Unicast instead of Multicast

caution: unicast support has only been added to Keepalived since version 1.2.8

By default Keepalived will use multicast packets to determine failover conditions. However, in many cloud environments it is not possible to use multicast because of network restrictions. Keepalived can be configured to use unicast in such environments:

  keepalived::vrrp::instance { 'VI_50':
    interface         => 'eth1',
    state             => 'BACKUP',
    virtual_router_id => '50',
    priority          => '100',
    auth_type         => 'PASS',
    auth_pass         => 'secret',
    virtual_ipaddress => '10.0.0.1/29',
    track_script      => 'check_nginx',
    unicast_source_ip => $::ipaddress_eth1,
    unicast_peers     => ['10.0.0.1', '10.0.0.2']
  }

The 'unicast_source_ip' parameter is optional as Keepalived will bind to the specified interface by default. The 'unicast_peers' parameter contains an array of ip addresses that correspond to the failover nodes.

Unit testing

Plain RSpec:

$ rake spec

Using bundle:

$ bundle exec rake spec

Test against a specific Puppet or Facter version:

$ PUPPET_VERSION=3.2.1  bundle update && bundle exec rake spec
$ PUPPET_VERSION=2.7.19 bundle update && bundle exec rake spec
$ FACTER_VERSION=1.6.8  bundle update && bundle exec rake spec