Skip to content

Insight and analysis of technology and business strategy

Making it easier to graph your infrastructure’s performance data

Today I would like to share a story with you about the development of a Puppet module for Grafana and its release on the Puppet Forge. But first, I'd like to provide some context so you can understand where Grafana fits in and why I feel this is important. Those of you that know me have likely heard me talk about the importance of data-driven decision making, and more specifically some of the tools that can be used to help enable individuals to make smart decisions about their IT infrastructure. A common approach is to deploy a graphing system such as Graphite, which stores performance data about your infrastructure to aide you in performing a number of functions including problem diagnosis, performance trending, capacity planning, and data analytics. If you are unfamiliar with the software, I’ll briefly describe its architecture. Graphite consists of a daemon, called carbon, which listens for time series data and writes it to a fixed-size database called whisper. It also provides a web application to expose the data and allow the user to create and display graphs on demand using a powerful API. While Graphite does a good job of storing time series data and providing a rich API for visualizing it, one of the things it does not really focus on is providing a dashboard for the data. Thankfully we have Grafana to fill this role and it happens to do it quite well. If you have ever worked with the ELK stack ( Elasticsearch, Logstash, and Kibana) before, Grafana’s interface should be familiar to you, as it is based on Kibana. It is a frontend for Graphite or InfluxDB, and runs as a client side application in your browser. Its only (optional) external dependency is Elasticsearch, as it can use it to store, load and search for dashboards. Below are some of Grafana’s most notable features (see its feature highlights for a more comprehensive list):
  • Dashboard search
  • Templated dashboards
  • Save / load from Elasticsearch and / or JSON file
  • Quickly add functions (search, typeahead)
  • Direct link to Graphite function documentation
  • Graph annotation
  • Multiple Graphite or InfluxDB data sources
  • Ability to switch between data sources
  • Show graphs from different data sources on the same dashboard
We like to make use of IT automation software whenever possible to deploy tools for our clients. Most tools already have Puppet modules or Chef cookbooks available for them, including the other components of the graphing system: Graphite itself, and a great Python-based collector named Diamond. Grafana, however, had no Puppet module available so I decided to rectify the situation by creating one and publishing it to the Puppet Forge. The module would be pretty simple: all that is required is to download and extract Grafana into an installation directory, and ensure appropriate values for the Elasticsearch, Graphite and InfluxDB servers / data sources are inserted into its configuration. I decided to offload the work of downloading and extracting the software to another module, namely gini/archive. And managing the configuration file, config.js, would be done with a combination of module parameters and ERB template. The only real complication arose when it came time to test serving Grafana with a web server such as Apache or Nginx. I decided not to have my module manage the web server in any way, so I would leverage Puppet Labs’ own Apache module for this purpose. My test environment consisted of a CentOS virtual machine provisioned by Vagrant and Puppet, with Graphite and Grafana on the same server. I decided to use Daniel Werdermann’s module to deploy Graphite on my virtual machine as it had worked well for me in the past. I quickly ran into problems with duplicate resources, however, due to the Graphite module managing Apache for creation of its virtual host etc. I moved to separate virtual machines for Graphite and Grafana, and that made my life easier. If you do decide to run both pieces of software on the same server, and are also using Daniel’s module, you can work around the problem by setting gr_web_server to ‘none’ like this:
class { 'graphite':
  gr_web_server => 'none',
  gr_web_cors_allow_from_all => true,
Since my module does not manage Apache (or Nginx), it is necessary to add something like the following to your node’s manifest to create a virtual host for Grafana:
# Grafana is to be served by Apache
 class { 'apache':
  default_vhost => false,
 # Create Apache virtual host
 apache::vhost { '':
  servername => '',
  port => 80,
  docroot => '/opt/grafana',
  error_log_file => 'grafana-error.log',
  access_log_file => 'grafana-access.log',
  directories => [
  path => '/opt/grafana',
  options => [ 'None' ],
  allow => 'from All',
  allow_override => [ 'None' ],
  order => 'Allow,Deny',
And the Grafana declaration itself:
class { 'grafana':
  elasticsearch_host => '',
  graphite_host => '',
Now that my module was working, it was time to publish it to the Puppet Forge. I converted my Modulefile to metadata.json, added a .travis.yml file to my repository and enabled integration with Travis CI, built the module and uploaded it to the Forge. Since its initial release, I have updated the module to deploy Grafana version 1.6.1 by default, including updating the content of the config.js ERB template, and have added support for InfluxDB. I am pretty happy with the module and hope that you find it useful. I do have plans to add more capabilities to the module, including support of more of Grafana’s configuration file settings, having the module manage the web server’s configuration similar to how Daniel’s module does it, and adding a stronger test suite so I can ensure compatibility with more operating systems and Ruby / Puppet combinations. I welcome any questions, suggestions, bug reports and / or pull requests you may have. Thanks for your time and interest! Project page: Puppet Forge URL:

Top Categories

  • There are no suggestions because the search field is empty.

Tell us how we can help!