Deploy to Kubernetes

Deploy a Kubernetes resource, as well as optionally a Kubernetes secret containing your Docker registry credentials. This secret containing the Docker registry credentials is always called “acr-auth”. Also can create a Kubernetes secret of any secrets available in your cloud vault that match the application name. This Kubernetes secret will be given the name: {application_name}-secret. All cloud vault secrets will be stored in a key-value form in this single Kubernetes Secret. If any of the Kubernetes resources already exists, this step will update them where appropriate (similar to what the kubectl apply -f command will do. In some cases, you may wish to restart a Kubernetes resource, even if the Kubernetes yaml configuration has not changed. An example of this is if you build a new Docker image, with the same tag. The default Kubernetes behaviour is to not restart the resource. Takeoff allows you to override this behaviour if so desired.

This task is usually used in combination with Build Docker Image (assuming your Kubernetes config references the image that is built)

Deployment

Add the following task to deployment.yaml:

- task: deploy_to_kubernetes
  kubernetes_config_path: my_kubernetes_config.yml.j2

This should be after the build_docker_image task if used together.

An example of kubernetes_config_path.yml.j2

apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: my-app
spec:
  template:
    metadata:
      labels:
        app: my-app
    spec:
      containers:
      - name: my-app
        image: my-docker-image:
        imagePullPolicy: 
        ports:
        - containerPort: 8443
        env:
        - name: SOME_SECRET
          valueFrom:
            secretKeyRef:
              name: my-app-secret
              key: some-secret
      imagePullSecrets:
      - name: acr-auth
---
apiVersion: v1
kind: Service
metadata:
  name: my_service
spec:
  ports:
  - port: 443
    protocol: TCP
    targetPort: 8443
  selector:
    app: my_app
  type: LoadBalancer
---
apiVersion: networking.k8s.io/v1beta1
kind: Ingress
metadata:
  name: my-ingress
  annotations:
    nginx.ingress.kubernetes.io/rewrite-target: /
spec:
  rules:
  - host: 
    http:
      paths:
      - path: /foo
        backend:
          serviceName: service1
          servicePort: 4200

An explanation for the Jinja templated values. These values get resolved automatically during deployment.

Other templated variables can be filled in two ways:

• Via your cloud keyvault, such as `` is a reference to a cloud vault key. The task will pull all secrets from the cloud vault prefixed with you application name and resolve them in the template. For the example above, if your application name is myapp, then a secret in your cloud vault must be myapp-secret-pull-policy or myapp-secret_pull_policy. The prefix gets removed by Takeoff and key secret_pull_policy with it’s value will be passed into the template. Hyphens - get normalized to underscores _.
• Via the custom_values configuration option specified in deployment.yml. Here, you are expected to set any custom key-value pairs, per environment. An example is shown below. In the above Kubernetes configuration, the `` key is filled by a custom value passed in via deployment.yml.

Takeoff config

Make sure .takeoff/config.yml contains the following keys:

azure:
  kubernetes_naming: "my_kubernetes{env}"
  keyvault_keys:
    container_registry:
      username: "registry-username"
      password: "registry-password"
      registry: "registry-server"

Default fields available for rendering

The following fields are available while rendering your Kubernetes yaml template:

| field | description | —– | ———– | docker_tag | the tag assigned to the docker image (based on environment configuration)| | application_name | the application name (used throughout takeoff) | | env | the deployment environment (used throughout takeoff) | | build_env | the contents of the build-runner’s environment variables |

Examples

Minimum Takeoff deployment configuration example to deploy Kubernetes resources. This will not create image pull secrets:

steps:
- task: deploy_to_kubernetes
  kubernetes_config_path: my_kubernetes_config.yml.j2
  image_pull_secret: 
    create: False

Extended configuration example, where we have explicitly disabled the creation of kubernetes secrets by Takeoff. In this case, we also want to restart the resources, even if their Kubernetes yaml config is unchanged. It will also create image pull secrets in namespace default with name registry-auth. We also pass in a custom url value per environment in this example. In this case, we’re using the default environment naming that Takeoff itself uses too. Please take a look at the environment docs for more information on how to define your own environment names.

steps:
- task: deploy_to_kubernetes
  kubernetes_config_path: my_kubernetes_config.yml.j2
  image_pull_secret: 
    create: True
  restart_unchanged_resources: true
  wait_for_rollout:
    resource_name: "deployment/my_app"
    resource_namespace: "my_namespace"
  custom_values:
    dev:
      url: 'dev-url-here-being-buggy'
    acp:
      url: 'acp-url-here-being-awesome'
    prd:
      url: 'prd-url-here-being-glorious'

Waiting for successful rollout

In the extended example above, you can see the wait_for_rollout parameter. This tells Takeoff that it should wait until the specified resource is rolled out successfully. If it is not rolled out successfully, or is not rolled out successfully quickly enough (e.g. within the default Kubernetes timeout), the task will fail. The failure of this task will trigger Kubernetes to rollback the change and revert to a previous revision that did work.

There are few things to note regarding this option:

1. It may result in unexpected behaviour if there are several CI jobs running simultaneously that apply changes to the specified Kubernetes resource. It is possible that the wrong version is awaited. This is unlikely, but not impossible.
2. Note that the resource_name must be prefixed with the resource_type. If this is not the case, a schema validation error will be thrown. Also note that not all resources are “awaitable”. For a full list of resources that can be awaited, we refer you to the Kubernetes documentation.

Takeoff Context

Eventhub producer policy secrets and consumer group secrets from configure_eventhub are available during this task. This makes it possible for the configuration below to inject the secrets into my_kubernetes_config.yml.j2:

steps:
  - task: configure_eventhub
    create_producer_policies:
      - eventhub_entity_naming: entity1
      - eventhub_entity_naming: entity2
    create_consumer_groups:
      - eventhub_entity_naming: entity3
  - task: deploy_to_kubernetes
    kubernetes_config_path: my_kubernetes_config.yml.j2

with my_kubernetes_config.yml.j2

apiVersion: v1
kind: Secret
metadata:
  name: armada-connections
data:
  entity1-producer-secret: 
  entity2-producer-secret: 
  entity3-consumer-secret: 

The jinja variables entity1_connection_string and entity2_connection_string are named by your eventhub_entity_naming in create_producer_policies, posfixed with connection_string.

Base64 encoding of secrets

Kubernetes requires the values of secrets to be base64 encoded. Takeoff enables this, and by default takes the following approach:

• Any values that are inserted into your Kubernetes template that originate from the Keyvault are assumed to be confidential. As such, Takeoff will ensure that these values are base64 encoded, as they should only be used in Kubernetes secrets.
• Any values that are inserted into your Kubernetes template via the custom values support (i.e. that are supplied in Takeoff’s deployment.yml) are assumed to not be confidential. These values will therefore be inserted into the template in plain text.

We believe these two assumptions are reasonable, and recommend you do not deviate from this. If, for some reason, you want/need to deviate from this, you can by using the following two filters in your Jinja template:

• b64_encode: apply base64 encoding. Example usage: ``
• b64_decode: apply base64 decoding. Example usage: ``

TL;DR: secrets are base64 encoded by default, custom_values are not. You can deviate from this using the jinja2 filters if you like.