telegraf

Table of Contents

1. Overview
2. Setup
3. Usage
4. Limitations
5. Development

Overview

A reasonably simple yet flexible Puppet module to manage configuration of InfluxData's Telegraf metrics collection agent.

Setup

There's a couple of fairly standard dependencies for this module, as follows:

• https://github.com/puppetlabs/puppetlabs-stdlib
• https://github.com/puppetlabs/puppetlabs-apt (on Debian / Ubuntu)

NB: On some apt-based distributions you'll need to ensure you have support for TLS-enabled repos in place. This can be achieved by installing the apt-transport-https package.

Usage

Telegraf's configuration is split into four main sections - global tags, options specific to the agent, input plugins, and output plugins. The documentation for these sections is here, and this module aims to be flexible enough to handle configuration of any of these stanzas.

To get started, Telegraf can be installed with a very basic configuration by just including the class:

include ::telegraf

However, to customise your configuration you'll want to do something like the following:

class { '::telegraf':
    hostname => $::hostname,
    outputs  => {
        'influxdb' => {
            'urls'     => [ "http://influxdb0.${::domain}:8086", "http://influxdb1.${::domain}:8086" ],
            'database' => 'telegraf',
            'username' => 'telegraf',
            'password' => 'metricsmetricsmetrics',
            }
        },
    inputs   => {
        'cpu' => {
            'percpu'   => true,
            'totalcpu' => true,
        },
    }
}

Or here's a Hiera-based example (which is the recommended approach):

---
telegraf::global_tags:
  role: "%{::role}"
  hostgroup: "%{::hostgroup}"
  domain: "%{::domain}"
telegraf::inputs:
  cpu:
    percpu: true
    totalcpu: true
  mem:
  io:
  net:
  disk:
  swap:
  system:
telegraf::outputs:
  influxdb:
    urls:
      - "http://influxdb0.%{::domain}:8086"
      - "http://influxdb1.%{::domain}:8086"
    database: 'influxdb'
    username: 'telegraf'
    password: 'telegraf'

telegraf::inputs accepts a hash of any inputs that you'd like to configure. However, you can also optionally define individual inputs using the telegraf::input type - this suits installations where, for example, a core module sets the defaults and other modules import it.

Example 1:

telegraf::input { 'my_exec':
  plugin_type => 'exec',
  options     => {
    'commands'    => ['/usr/local/bin/my_input.py',],
    'name_suffix' => '_my_input',
    'data_format' => 'json',
  },
  require     => File['/usr/local/bin/my_input.py'],
}

Will create the file /etc/telegraf/telegraf.d/my_exec.conf:

[[inputs.exec]]
  commands = ['/usr/local/bin/my_input.py']
  name_suffix = '_my_input'
  data_format = 'json'

Example 2:

telegraf::input { 'influxdb-dc':
  plugin_type => 'influxdb',
  options     => {
    'urls' => ['http://remote-dc:8086',],
  },
}

Will create the file /etc/telegraf/telegraf.d/influxdb-dc.conf:

[[inputs.influxdb]]
  urls = ["http://remote-dc:8086"]

Example 3:

telegraf::input { 'my_snmp':
  plugin_type => 'snmp',
  options     => {
    'interval' => '60s',
  },
  sections    => {
    'snmp.host' => {
      'address'   => 'snmp_host1:161',
      'community' => 'read_only',
      'version'   => 2,
      'get_oids'  => ['1.3.6.1.2.1.1.5',],
    },
  },
}

Will create the file /etc/telegraf/telegraf.d/snmp.conf:

[[inputs.snmp]]
  interval = "60s"

[[inputs.snmp.host]]
  address = "snmp_host1:161"
  community = "read_only"
  version = 2
  get_oids = ["1.3.6.1.2.1.1.5"]

Hierarchical configuration from multiple files

Hiera YAML and JSON backends support deep hash merging which is needed for inheriting configuration from multiple files.

First of all, make sure that the deep_merge gem is installed on your Puppet Master.

An example of hiera.yaml:

---
:hierarchy:
    - "roles/%{role}"
    - "type/%{virtual}"
    - "domain/%{domain}"
    - "os/%{osfamily}"
    - "common"
:backends:
    - yaml
:yaml:
    :datadir: /etc/puppet/hiera
:merge_behavior: deeper

Then you can define configuration shared for all physical servers and place it into type/physical.yaml:

telegraf::inputs:
  cpu:
    percpu: true
    totalcpu: true
  mem:
  io:
  net:
  disk:

Specific roles will include some extra plugins, e.g. role/frontend.yaml:

telegraf::inputs:
  nginx:
    urls: ["http://localhost/server_status"]

Limitations

This module has been developed and tested against:

• Ubuntu 14.04
• Debian 8
• CentOS / RHEL 6
• CentOS / RHEL 7

Support for other distributions / operating systems is planned. Feel free to assist with development in this regard!

The configuration generated with this module is only compatible with newer releases of Telegraf, i.e 0.11.x. It won't work with the 0.2.x series.

Development

Please fork this repository, hack away on your branch, run the tests:

$ bundle exec rake test acceptance

And then submit a pull request. Succinct, well-described and atomic commits preferred.