README.md

consul-cookbook

Application cookbook which installs and configures Consul.

Consul is a tool for discovering and configuring services within your infrastructure. This is an application cookbook which takes a simplified approach to configuring and installing Consul. Additionally, it provides Chef primitives for more advanced configuration.

Basic Usage

For most infrastructure we suggest first starting with the default recipe. This installs and configures Consul from the latest supported release. It is also what is used to certify platform support through the use of our integration tests.

This cookbook provides node attributes which are used to fine tune the default recipe which installs and configures Consul. These values are passed directly into the Chef resource/providers which are exposed for more advanced configuration.

Out of the box the following platforms are certified to work and are tested using our Test Kitchen configuration. Additional platforms may work, but your mileage may vary.

• CentOS (RHEL) 6.6, 7.1
• Ubuntu 12.04, 14.04
• Windows

Client

Out of the box the default recipe installs and configures the Consul agent to run as a service in client mode. The intent here is that your infrastructure already has a [quorum of servers][13]. In order to configure Consul to connect to your cluster you would supply an array of addresses for the Consul agent to join. This would be done in your wrapper cookbook:

node.default['consul']['config']['start_join'] = %w{c1.internal.corporate.com c2.internal.corporate.com c3.internal.corporate.com}

Server

This cookbook is designed to allow for the flexibility to bootstrap a new cluster. The best way to do this is through the use of a wrapper cookbook which tunes specific node attributes for a production server deployment.

The [Consul cluster cookbook][14] is provided as an example.

Advanced Usage

As explained above this cookbook provides Chef primitives in the form of resource/provider to further manage the install and configuration of Consul. These primitives are what is used in the default recipe, and should be used in your own wrapper cookbooks for more advanced configurations.

Configuration

It is very important to understand that each resource/provider has defaults for some properties. Any changes to a resource's default properties may need to be also changed in other resources. The best example is the Consul configuration directory.

In the example below we're going to change the configuration file from the default (/etc/consul.json) to one that may be on a special volume. It is obvious that we need to change the path where consul_config writes its file to, but it is less obvious that this needs to be passed into consul_service.

Inside of a recipe in your wrapper cookbook you'll want to do something like the following block of code. It uses the validated input from the configuration resource and passes it into the service resource. This ensures that we're using the same data.

config = consul_config '/data/consul/default.json'
consul_service 'consul' do
  config_file config.path
end

Watches/Definitions

In order to provide an idempotent implementation of Consul watches and definitions. We write these out as a separate configuration file in the JSON file format. The provider for both of these resources are identical in functionality.

Below is an example of writing a Consul service definition for the master instance of Redis. We pass in several parameters and tell the resource to notify the proper instance of the Consul service to reload.

consul_definition 'redis' do
  type 'service'
  parameters(tags: %w{master}, address: '127.0.0.1', port: 6379, interval: '30s')
  notifies :reload, 'consul_service[consul]', :delayed
end

A check definition can easily be added as well. You simply have to change the type and pass in the correct parameters. The definition below checks memory utilization using a script on a ten second interval.

consul_definition 'mem-util' do
  type 'check'
  parameters(script: '/usr/local/bin/check_mem.py', interval: '10s')
  notifies :reload, 'consul_service[consul]', :delayed
end

Finally, a watch is created below to tell the agent to monitor to see if an application has been deployed. Once that application is deployed a script is run locally. This can be used, for example, as a lazy way to clear a HTTP disk cache.

consul_watch 'app-deploy' do
  type 'event'
  parameters(handler: '/usr/local/bin/clear-disk-cache.sh')
  notifies :reload, 'consul_service[consul]', :delayed
end

A keen eye would notice that we are delaying the reload of the Consul service instance. The reason we do this is to minimize the number of times we need to tell Consul to actually reload configurations. If there are several definitions this may save a little time off your Chef run.

Execute

The command-line agent provides a mechanism to facilitate remote execution. For example, this can be used to run the uptime command across your fleet of nodes which are hosting a particular API service.

consul_execute 'uptime' do
  options(service: 'api')
end

All of the options available on the command-line can be passed into the resource. This could potentially be a very dangerous operation. You should absolutely understand what you are doing. By the nature of this command it is impossible for it to be idempotent.

UI

consul_ui resource can be used to download and extract the consul web UI. It can be done with a block like this:

consul_ui 'consul-ui' do
  owner node['consul']['service_user']
  group node['consul']['service_group']
  version node['consul']['version']
end

Assuming consul version 0.5.2 above block would create /srv/consul-ui/0.5.2 and symlink /srv/consul-ui/current.

It does not change agent's configuration by itself. consul_config resource should be modified explicitly in order to host the web page.

consul_config 'consul' do
  ...
  ui_dir '/srv/consul-ui/current/dist'
end

This is optional, because consul UI can be hosted by any web server.

[13]: [14]: https://github.com/johnbellone/consul-cluster-cookbook