Kubernetes Pipeliner

This tool is used to generate pipeline JSON for Spinnaker from Kubernetes manifest files.

The basic premise is that Kubernetes already has a well defined standard for how to define cluster resources (ReplicaSets, Deployments, etc), but there's no way to glue them into Spinnaker pipeline stages. That's what this tool aims to provide.

Using a pipeline configuration YAML file, you can define stages and reference Kubernetes manifest definitions to help fill in:

• The environment variables
• The container image
• The command and arguments
• Ports and load balancers

Example Pipeline

This is a lengthy example of how a pipeline YAML looks.

name: Example Deployment
application: example

disableConcurrentExecutions: true
keepQueuedPipelines: true
description: This pipeline deploys some sweet code

notifications:
  - address: "#launchpad"
    type: "slack"
    when:
      - pipeline.complete
      - pipeline.failed
    message:
      pipeline.complete: |
        The stage has completed!
      pipeline.failed: |
        The stage has failed!

parameters:
  - name: "random"
    description: "random description"
    required: true
    default: "hello-world"

triggers:
- jenkins:
    job: "Example/job/master"
    master: "jenkins"
    propertyFile: "build.properties"
    enabled: true
- webhook:
    source: "random-string"
    enabled: true
imageDescriptions:
  - name: main-image
    account: "namely-registry"
    image_id: "${ trigger.properties['docker_image'] }"
    registry: "registry.namely.land"
    repository: "namely/example-all-day"
    tag: "${ trigger.properties['docker_tag'] }"
    organization: "namely"
  - name: second-image
    account: "namely-registry"
    image_id: "${ trigger.properties['docker_image'] }"
    registry: "registry.namely.land"
    repository: "namely/second-repo-name"
    tag: "${ trigger.properties['docker_tag'] }"
    organization: "namely"
stages:
- account: int-k8s
  name: "Migrate INT"
  runJob:
    manifestFile: test-deployment.yml
    imageDescriptions:
      - name: main-image
        containerName: example
    container: # override default command defined in the manifest
      command:
        - bundle
        - exec
        - rake
        - db:migrate
  notifications:
    - address: "#launchpad"
      type: "slack"
      when:
        - stage.complete
        - stage.failed
      message:
        stage.complete: |
          The stage has completed!
        stage.failed: |
          The stage has failed!
- account: int-k8s
  name: "Deploy to INT"
  deploy:
    groups:
      - manifestFile: test-deployment.yml
        maxRemainingASGS: 2 # total amount of replica sets to keep around afterwards before deleting
        scaleDown: true
        stack: web # primarily for labeling purposes on the created resources
        details: helloworld
        strategy: redblack
        targetSize: 2 # this is the total amount of replicas for the deployment
        containerOverrides: {}
        imageDescriptions:
          - name: main-image
            containerName: example
          - name: second-image
            containerName: init-example
        loadBalancers:
          - "test"
- account: int-k8s
  name: "Proceed to Staging?"
  refId: "3"
  reliesOn:
    - "2"
  manualJudgement:
    timeoutHours: 48
    failPipeline: true
    instructions: |
      If this stage has completed QA, press proceed.

Installation

If you have a Go environment installed and configured, you can use go get to install the latest package of this project:

$ go install github.com/namely/k8s-pipeliner/cmd/k8s-pipeliner

To use it:

$ cd your-project
$ k8s-pipeliner create pipeline.yml

If you want a pretty view and have JQ installed, you can do:

$ k8s-pipeliner create pipeline.yml | jq .

To copy the result to your clipboard and you're on a Mac, you can do:

$ k8s-pipeliner create --linear pipeline.yml | pbcopy

Upgrade k8s-pipeliner

Pull the latest from master branch and run

make install

Adding the Pipeline JSON

Once you've copied the resulting JSON from the pipeline configuration, you can go modify an already created Pipeline by clicking "Pipeline Actions" -> "Edit as JSON".

Paste the JSON, and then in the bottom right of the screen click "Save".

Schema

Here are the independent pieces of schema for pipeline.yml that you can use. You can also take a look at the Config Definitions.

Triggers

We currently support 2 types of triggers in k8s-pipeliner, webhooks and jenkins.

Webhooks

triggers:
- webhook:
    source: "random-string"
    enabled: true

The "source" field in webhooks is the endpoint that you need to hit in order to kick off Spinnaker. If you're running the gate API at "gate-api.example.com", your webhook endpoint from this configuration would be https://gate-api.example.com/webhooks/webhook/random-string (random-string is our source value)

Jenkins

Spinnaker also supports triggering off of Jenkins jobs completing, to use this trigger include a jenkins field:

triggers:
- jenkins:
    job: "Example/job/master"
    master: "jenkins"
    propertyFile: "build.properties" # optional
    enabled: true

Manual Judgement

If you want to have a manual judgement in your pipeline, you can define a manualJudgement step within the stages array:

stages:
- name: "Continue Deploy?"
  manualJudgement:
    failPipeline: true
    instructions: |
      Once you're confident with this deploy, please approve it to continue.

• If failPipeline is set to true, the manual judgement must be approved for the rest of the pipeline to continue.
• The instructions are displayed within the UI when the pipeline is stalled waiting for a manual judgement. This is useful for whoever is providing the manual judgement to have context.

Run A Job

A Job is a step in a pipeline that runs a one off task. A good example might be running a database migration before rolling out a piece of code.

stages:
- name: "Run Migrations"
  runJob:
    manifestFile: manifests/deployment.yml
    container:
      command:
        - bundle
        - exec
      args:
       - rake
       - db:migrate

• manifestFile is used to generate the majority of the stage JSON for running the container. Things like environment, volumes, commands, etc are all stored within this Kubernetes Manifest file.
• container key is used to overwrite some of the values that are provided in the manifestFile. For example, if you want to run a migration script that is provided in the container instead of the default rails server, this is where you'd define it.
• command portion of the container override overwrites the command portion of the container being run in the job.
• args portion of the container override overwrites the args portion of the container being run in the job.

Deploying Groups

A Deploy stage is used for running new server groups. You can use this stage to deploy several groups in-tandem. This is useful if you're deploying the same container for different application needs. IE: One is a consumer and another is a publisher.

imageDescriptions:
  - name: main-image
    account: "namely-registry"
    image_id: "${ trigger.properties['docker_image'] }"
    registry: "registry.namely.land"
    repository: "namely/example-all-day"
    tag: "${ trigger.properties['docker_tag'] }"
    organization: "namely"

stages:
- name: "Deploy"
  deploy:
    groups:
    - manifestFile: manifests/deployment.yml
      maxRemainingASGS: 2
      scaleDown: true
      stack: web
      details: genpop
      strategy: redblack
      targetSize: 10
      loadBalancers:
        - namely
      imageDescriptions:
        - name: main-image
          containerName: my-container

• manifestFile is used to generate the majority of the stage JSON for running the container. Things like environment, volumes, commands, etc are all stored within this Kubernetes Manifest file.
• maxRemainingASGS determines how many ReplicaSets Spinnaker will keep around after a deploy occurs. If using redblack strategy you need at least 2. This is used for rolling back deploys.
• scaleDown scales down the previous server group after a deploy. If you want traffic to be routed to both deployments set this to false
• stack is concatenated to the application name when deploying. So application-stack would be a result. CANNOT have dashes.
• details is concatenated to the application name and stack when deploying. So application-stack-detail would be a result. This can have multiple dashes.
• strategy is used to determine which strategy Spinnaker should use when deploying this new group.
• targetSize is the amount of replicas to be deployed to the Kubernetes cluster. This is not taken from the deployment manifest file
• loadBalancers are the Spinnaker load balancers to be attached to this deployment. An array of strings. These will need to be defined inside of Spinnaker before a deploy to work.

Image Descriptions

imageDescriptions are a top level key in your YAML that define a Docker image to be deployed. You can use Spinnaker expression syntax in these fields to add dynamic images in your deployment pipeline.

The name key on the image description is then referenced in your groups of a deploy stage. You can see this here:

imageDescriptions:
  - name: main-image
    containerName: my-container

What k8s-pipeliner does is it looks into the manifest you've supplied, finds the container with the name "my-container", and includes the image description for the Spinnaker JSON that is rendered for it. This allows you to specify multiple containers in your pods and be able to swap out the images based on dynamic values for them.

Parameter Support

This tool also supports the ability to include parameters in your pipeline definitions:

parameters:
  - name: "random"
    description: "random description"
    required: true
    default: "hello-world"

This configures your pipeline to have parameters in the UI / enable pipeline expressions.

Evaluate Variables Stage

The evaluate variables stage allows you to evaluate complex expressions and use them throughout your pipeline:

- name: "evaluate variables"
  variables:
    - key: "my-fun-key"
      value: ${my-complex-expression}

You can reference the variable by the key name:

- name: "Some other stage"
  someField: ${ my-fun-key }

Configurator

Files under the configuratorFiles section are expected to be in the k8s-configurator format. These will be run through k8s-configurator to generate the environment-specific manifest. By default, the environment used by k8s-configurator will be determined by the account used in this stage. However, you may set the optional env property for configuratorFiles to override this.

stages:
- account: ops-k8s
  name: "Deploy"
  deployEmbeddedManifests:
    moniker:
      app: app
      cluster: cluster
      detail: detail
      stack: stack
    files:
      - file: manifests/deployment.yml
      - file: manifests/service.yml
      - file: manifests/migrate-job.yml
    configuratorFiles:
      - file: test-configurator.yml
        env: superOps
    containerOverrides:
      name: "container-name"
      resources:
        requests:
          memory: "100"
          cpu: "200"

All of these files will be composed into a single stage deployment into the given account. This means you can deploy services and deployments in tandem together.