The Twelve-Factor App says that web applications should retrieve their configuration from environmental variables. This practice has been rapidly adopted by modern PaaS services to enable simple configuration changes.
With Consul, it is simple to bring this practice to your own datacenters. If you use a PaaS for some aspects of your infrastructure but not all of it, Consul is a great way to centralize configuration data.
In this post, we show how Consul and envconsul can be used to set configuration values and trigger automatic restarts on configuration changes, all without any modifications to your applications.
According to the Twelve-Factor app, environmental variables should be used for web application configuration. They have a number of benefits when compared against configuration files or mechanisms such as Java System Properties:
Environmental variables are a language- and OS-agnostic standard.
Environmental variables are more difficult to accidentally commit to a code repo.
Environmental variables are easy to change between environments such as development, staging, QA.
Environmental variables are easy to set and update regardless of deploy.
Complete PaaS solutions such as Heroku expose helpful APIs to automatically set/get environmental variables for an application.
When deploying applications manually, the story has historically been a bit more complicated. With Consul, it is just as easy to set and read configurations for developers, and easy to support and maintain as an operations engineer.
It is clear to see how to set and retrieve configuration data, but it isn't clear how this configuration data can become environmental variables for an application. envconsul is a lightweight solution that solves this problem.
With envconsul, environmental variables are stored in Consul KV under some prefix (separated by "/"). For example, to configure our service "foo" we might store configuration like this:
$ curl -X PUT -d 'false' http://localhost:8500/v1/kv/foo/enabled true
This stores the value
false in the key
Then, with envconsul, we can turn these keys into environmental variables:
$ envconsul foo env ENABLED=false
envconsul is a very UNIX-friendly application. It takes two mandatory arguments: a KV prefix to find data and then another application to run along with its arguments. In the above example, we tell envconsul that our configuration is under the prefix
foo and we want to run the application
env, which simply outputs the environmental variables.
In the result, you can clearly see that
ENABLED has been set to
false, just like we set in Consul KV.
If you change
env to your application, then the environmental variables will be exposed to that application. For example, to run a rails server you might do the following. Note that in a real production scenario, you probably aren't running the Rails built-in server directly, but it makes for a good example:
$ envconsul foo bin/rails server ...
With a PaaS, your application is automatically restarted when you change any configuration. We can achieve the same effect with Consul and Envconsul with minimal effort.
By adding the
-reload flag to envconsul, envconsul will terminate (SIGTERM) and restart your application whenever a configuration key is added, removed, or changed:
$ envconsul -reload foo bin/rails server ...
The Consul HTTP API supports long polling for changes in the KV for a given prefix. Envconsul uses this to efficiently detect changes in the KV as soon as they happen.
Using Consul and envconsul for application configuration can bring the ease of PaaS-like application configuration into your own native environments.
For developers, they are able to set configuration without talking to operations engineers or redeploying the application.
For operations, Consul provides a uniform solution for service discovery and configuration across the entire infrastructure. Operations can sleep well knowing that Consul automatically replicates this data and stores it on disk for easy backup.
This is just one of many ways Consul can improve your datacenter. We didn't cover the benefits of Consul's service discovery at all, nor did we talk about the technical details about how Consul keeps your data safe.
You can even use Consul KV and envconsul to sync configuration to or from your datacenter and a 3rd party PaaS! For large organizations with a hybrid cloud environment, this allows centralization of configuration that was difficult to achieve before.
If you'd like to learn more about Consul, please visit the Consul website. We plan on doing more blog posts on applications of Consul as well some deep dive technical blog posts on how Consul works.
Consul 1.9 introduces new service mesh visualization tools that help users troubleshoot their mesh and view key metrics.
The HashiCorp Cloud Platform and its Consul service are now available for all to try in public beta.
Consul 1.9 adds more service mesh features including app-aware intentions, visualization enhancements, and custom resource definitions.