Kube::Templates

Simple template system for Kubernetes deployments.

We built this to handle deploying resque workers as containers. For each of our Rails apps we may have 50 or more resque workers running that code base, listening to different queues. This simplified the deployment by allowing us to provide a single Deployment template and then generate lots of deployment configurations with the right settings for the queues for the workers to listen to and the number of replicas.

Add this line to your application's Gemfile:

gem "kube-templates"

And then execute:

$ bundle

Or install it yourself as:

$ gem install kube-templates

Kube::Templates::Builder takes a template file and a configuration file, assigning values from the configuration file and emitting a collection of deployments for kubernetes to run.

Create a deployment template. In this file you can have template variables (e.g. ${QUEUES}) that will be replaced with values from your configuration. In the following we use three variables, ${NAME}, ${REPLICAS}, and ${QUEUES}.

# resque-template.yaml
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: resque-${NAME}
spec:
  replicas: ${REPLICAS}
  template:
    metadata:
      labels:
        service: rails
        app: my-app
        purpose: worker
        queues: ${QUEUES}
    spec:
      containers:
      - name: my-app
        image: us.gcr.io/project-id-1138/my-app-resque:latest
        env:
        - name: QUEUE
          value: ${QUEUES}

Create a configuration file that defines your workers. In the following file we define two types of workers. One, with four replicas, listens onthe "reports" queue. A second listens on the "process_priority" and "process" queues. This second worker will have two replicas set as that values it defined in the "defaults" section and it doesn't change it.

# resque-workers.yml
defaults:
  replicas: 2
workers:
  - queues: reports
    replicas: 4
    name: builder
  - queues: process_priority,process

Note that while we can define the NAME variable, as we did for the first worker, we don't need to and the Builder will automatically generate a name from the values provided for the worker. So the second worker would have a name "process-priority-process"

Use the resque-k8s command to create a series of YAML deployments for kubernetes to consume. You can pipe the output to less to see what it produces.

$ resque-k8s | less

To apply this run the kubectl apply command with the value from the configuration.

$ resque-k8s | kubectl apply -f -

This is a very simple implementation. There are a number of solutions out there that propose to do something similar in the scope of larger efforts.

• There is currently a proposal to support this in Kubernetes.

• kb8or includes something like this with many more continuous deployment features

• Kubernetes Helm does this with its charts.

After checking out the repo, run bin/setup to install dependencies. Then, run rake spec to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install.

To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and tags, and push the .gem file to rubygems.org.

Bug reports and pull requests are welcome on GitHub at https://github.com/keylimetoolbox/kube-templates.

The gem is available as open source under the terms of the MIT License.