Override the runner pod spec

Both TerraformRepository and TerraformLayer expose a spec.overrideRunnerSpec map field.

If the field is specified for a given TerraformRepository it will be applied by default to all TerraformLayer linked to it.

If the field is specified for a given TerraformLayer it will take precedence over the TerraformRepository configuration.

Info

• Maps (dictionaries): A deep merge is performed. Keys in TerraformLayer overwrite those in TerraformRepository, but unmatched keys are preserved.
• Arrays (lists): Arrays are not merged; they are fully replaced.

Available overrides are:

Fields
Affinity
Args
Command
ImagePullSecrets
Image
InitContainers
Tolerations
NodeSelector
ServiceAccountName
Resources
Env
EnvFrom
Volumes
VolumeMounts
Metadata.Annotations
Metadata.Labels
ExtraInitArgs
ExtraPlanArgs
ExtraApplyArgs

For instance with the following configuration, all the runner pods will have the specifications described inside the TerraformRepository:

apiVersion: config.terraform.padok.cloud/v1alpha1
kind: TerraformRepository
metadata:
  name: burrito
  namespace: burrito
spec:
  repository:
    url: https://github.com/padok-team/burrito
  terraform:
    enabled: true
  overrideRunnerSpec:
    imagePullSecrets:
    - name: ghcr-creds
    tolerations:
    - effect: NoSchedule
      key: burrito.io/production
      operator: Exists
    nodeSelector:
      production: "true"
    serviceAccountName: "production"
---
apiVersion: config.terraform.padok.cloud/v1alpha1
kind: TerraformLayer
metadata:
  name: random-pets
  namespace: burrito
spec:
  terraform:
    version: "1.3.1"
  path: "internal/e2e/testdata/random-pets"
  branch: "main"
  repository:
    name: burrito
    namespace: burrito

In the following case, nodeSelector will be merged and tolerations will be replaced:

apiVersion: config.terraform.padok.cloud/v1alpha1
kind: TerraformRepository
metadata:
  name: burrito
  namespace: burrito
spec:
  repository:
    url: https://github.com/padok-team/burrito
  terraform:
    enabled: true
  overrideRunnerSpec:
    imagePullSecrets:
    - name: ghcr-creds
    tolerations:
    - effect: NoExecute
      key: burrito.io/production
      operator: Exists
    nodeSelector:
      production: "true"
    serviceAccountName: "production"
---
apiVersion: config.terraform.padok.cloud/v1alpha1
kind: TerraformLayer
metadata:
  name: random-pets
  namespace: burrito
spec:
  terraform:
    version: "1.3.1"
  path: "internal/e2e/testdata/random-pets"
  branch: "main"
  repository:
    name: burrito
    namespace: burrito
  overrideRunnerSpec:
    tolerations:
    - effect: NoSchedule
      key: burrito.io/production
      operator: Exists
    nodeSelector: {}

Resulting in the following podSpec:

tolerations:
- effect: NoSchedule
  key: burrito.io/production
  operator: Exists
nodeSelector:
  production: "true"

Prepare runner configuration with initContainers

You may need to get refreshed tokens to access your APIs or prepare some dynamic configuration for Terraform. To do so, you can use initContainers.

Note

In order to allow access to files generated by your init container from your main container, you'll need to match the uid/gid from the runner image. You can check it here

apiVersion: config.terraform.padok.cloud/v1alpha1
kind: TerraformLayer
metadata:
  name: my-layer
spec:
  overrideRunnerSpec:
    volumes:
      - name: kubeconfig-volume
        emptyDir: {}
    volumeMounts:
      - name: kubeconfig-volume
        mountPath: /home/burrito/.kube
        readOnly: true
    initContainers:
      - name: init-runner
        image: registry.domain.com/image:tag
        securityContext:
          runAsNonRoot: true
          runAsUser: 65532  # match USER from Dockerfile
          runAsGroup: 65532 # match USER from Dockerfile
        command:
          - 'sh'
          - '-c'
          - 'generate-kube-config --output=/.kube/config'
        envFrom:
          - secretRef:
              name: kube-secrets
        volumeMounts:
          - mountPath: /.kube
            name: kubeconfig-volume