Possible to use Ansible variables in deployments?

[mikkeschiren]

I have a deployment which adds a cronjob to a deployment, and I want that cronjob to use variables from ansible - is that possible? Or could I in another way get variables from an operator?

Here is an example what I would like to do:

apiVersion: mthing.com/v1alpha1
kind: Foo
metadata:
  name: foo-com
  version: v0.1
  labels:
    app: web
---
apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: '{{ meta.name }}-cron' // This should be meta.name from  mthing.com/v1alpha1
spec:
  schedule: "*/4 * * * *"
  jobTemplate:
    spec:
      template:
        spec:
          containers:
          - name: '{{ meta.name }}-cron'
            image: gcr.io/maynard-io-public/kubernetes-s3-mysql-backup
            imagePullPolicy: Always
            env:
              - name: MYVAR
                value: name: '{{ meta.name }}-var'

When trying this, the operator errors out and complains about '{{ meta.name }} - which it thinks is a string.

[gearoidibm]

Same as I'm seeing in #1701 ?

[fabianvf]

Yes, any field in the spec of the custom resource being reconciled will be converted from camel to snake case and passed in to Ansible as an extra_var, and the meta field with name and namespace is passed in as well.

It looks like the sample you posted has a yaml formatting issue though,

value: name: '{{ meta.name }}-var'

Should be

value: '{{ meta.name }}-var'``

Additionally, make sure you are running the Ansible templating engine when reading from that file. The most common pattern is to use the template lookup plugin and pass it into the k8s module, like so:

- k8s:
    definition: "{{ lookup('template', 'cronjob.yaml.j2') }}"

[mikkeschiren]

Ah - I see! This really opens up for possibilities

Ok, I tried it like this (I know this will not work because of mix of kubernetes syntax and ansible, but how should you do it?)
(file: foo.yaml)

apiVersion: mthing.com/v1alpha1
kind: Foo
metadata:
  name: foo-com
  version: v0.1
  labels:
    app: web

---
- name: Cron
  k8s:
    definition: "{{ lookup('template', 'cron.yaml.j2') }}"

But when I am try to deploy the file:

kubectl deploy -f foo.yaml

I am getting

error: error validating "foo.yaml": error validating data: invalid object to validate

[flickerfly]

So I have a task (in roles/$name/tasks/main.yml) that looks like this:

- name: "Create {{ meta.name[0]|upper}}{{meta.name[1:] }} Service"
  k8s:
    definition: "{{ lookup('template', 'templates/service.yml.j2') }}"

and the template (in roles/$name/templates/services.yml.j2) looks like this:

apiVersion: v1
kind: Service
metadata:
  labels:
    app: '{{ meta.name }}'
  name: '{{ meta.name }}'
  namespace: '{{ meta.namespace }}'
spec:
  ports:
    - name: '{{ meta.name }}-http'
      port: '{{ httpport }}'
      protocol: TCP
      targetPort: '{{ httpport }}'
  selector:
    app: '{{ meta.name }}'
  sessionAffinity: None
  type: ClusterIP

The CR that it uses to gather the variables looks something like this and it added with kubectl create -f thisfile.yaml assuming the CRD has already been created. :

apiVersion: me.com/v1alpha1
kind: mykind
metadata:
  name: example-mykind
  namespace: mynamespace
spec:
  # Add fields here
  httpport: 8080