Centralized Docker container configuration with Tiller


Note : This page may contain outdated information and/or broken links; some of the formatting may be mangled due to the many different code-bases this site has been through in over 20 years; my opinions may have changed etc. etc.

If you’ve been using Docker for a while, you probably know that you can use Tiller to generate configuration files inside your containers. This means you can provide a single container for running in a variety of different environments (think of a web application that needs different database URIs and credentials depending on where it is run).

You can also use it to provide ‘parameterized’ containers - where you allow end-users to provide some aspects of the configuration. For example, the team over at StackStorm are using Tiller and the environment plugin to let users easily configure their containers. I’ve provided a variety of plugins to help with this, and have written about it previously.

There was always a catch, though. Until recently, Tiller obtained configuration from a ‘local’ source - either configuration files bundled in a container, or from environment variables passed in at container run time. This is probably fine for most situations, but did prevent a lot of interesting use-cases. However, the 0.6.x releases now support fetching configuration and templates from a variety of network sources. Currently (as of 0.6.1), you can retrieve data from either a ZooKeeper cluster or a HTTP server, with more services to be supported in later releases.

Update: Tiller now supports Consul!

This is a massive change as it now means you can have a central store for all your container configuration, and opens the door for a whole bunch of really cool possibilities. You could, for example, have a service that allows users to edit configuration files or set parameters through a web interface. Or you could plug Tiller into your automation stack so you can alter container configuration on the fly. You could even use Tiller “on the metal” to configure physical/VM hosts when they boot.

Of course, this does introduce an external dependancy on launching your containers, so you should ensure that before you implement this in a production setting you have carefully considered all your points of failure!

Enabling these plugins is straightforward - simply add them to common.yaml, e.g.

  - file
  - http

  - http
  - file

As with all Tiller plugins, the ordering is significant. In the above example, values from the HTTP data source will take precedence over YAML files, but templates loaded from files will take precedence over templates stored in HTTP. You should tweak this as appropriate for your environment.

You also do not need to enable both plugins; for example you may just want to retrieve values for your templates from a web server, but continue to use files to store your actual template content.

As the HTTP plugins can fetch everything (a list of templates, contents and values) from a webservice, if you accept the defaults your environment configuration blocks can literally now be reduced to :

    uri: 'http://tiller.example.com'

You can check out the documentation for these new plugins over at the Github project page. If you have any suggestions for improvements or feature requests, please feel free to open an issue or leave a comment below!