K8s-manifest Unit ¶

Info

This unit is deprecated. We suggest using the Kubernetes unit instead.

Applies Kubernetes resources from manifests.

Example:

- name: kubectl-test2
  type: k8s-manifest
  namespace: default
  create_namespaces: true
  path: ./manifests/
  apply_template: true
  kubeconfig: {{ output "this.kubeconfig.kubeconfig_path" }}
  kubectl_opts: "--wait=true"

Options ¶

force_apply - bool, optional. By default is false. If set to true, the unit will be applied when any dependent unit is changed.
namespace - optional. Corresponds to kubectl -n.
create_namespaces - bool, optional. By default is false. If set to true, cdev will create namespaces required for the unit (i.e. the namespaces listed in manifests and the one specified within the namespace option), in case they don't exist.
path - required, string. Indicates the resources that are to be applied: a file (in case of a file path), a directory recursively (in case of a directory path) or URL. In case of URL path the unit will download the resources by the link and then apply them.

Example of URL path:
```
- name: kubectl-test2
  type: k8s-manifest
  namespace: default
  path: https://git.io/vPieo
  kubeconfig: {{ output "this.kubeconfig.kubeconfig_path" }}
```
apply_template - bool. By default is set to true. See Templating usage below.
kubeconfig - optional. Specifies the path to a kubeconfig file. See How to get kubeconfig subsection below.
kubectl_opts - optional. Lists additional arguments of the kubectl command.

Templating usage ¶

As manifests are part of a stack template, they also maintain templating options. Specifying the apply_template option enables you to use templating in all Kubernetes manifests located with the specified path.

How to get kubeconfig ¶

There are several ways to get a kubeconfig from a cluster and pass it to the units that require it (for example, helm, K8s-manifest). The recommended way is to use the shell unit with the option force_apply. Here is an example of such unit:

- name: kubeconfig
  type: shell
  force_apply: true
  depends_on: this.k3s
  apply:
    commands:
      - aws s3 cp s3://{{ .variables.bucket }}/{{ .variables.cluster_name }}/kubeconfig /tmp/kubeconfig_{{ .variables.cluster_name }}
      - echo "kubeconfig_base64=$(cat /tmp/kubeconfig_{{ .variables.cluster_name }} | base64 -w 0)"
      - echo "kubeconfig_path=/tmp/kubeconfig_{{ .variables.cluster_name }}"
  outputs:
    type: separator
    separator: "="

In the example above, the K3s unit (the one referred to) will deploy a Kubernetes cluster in AWS and place a kubeconfig file in S3 bucket. The kubeconfig unit will download the kubeconfig from the storage and place it within the /tmp directory.

The kubeconfig can then be passed as an output to other units:

- name: cert-manager-issuer
  type: k8s-manifest
  path: ./cert-manager/issuer.yaml
  kubeconfig: {{ output "this.kubeconfig.kubeconfig_path" }}

An alternative (but not recommended) way is to create a yaml hook in a stack template that would take the required set of commands:

_: &getKubeconfig "rm -f ../kubeconfig_{{ .name }}; aws eks --region {{ .variables.region }} update-kubeconfig --name {{ .name }} --kubeconfig ../kubeconfig_{{ .name }}"

and execute it with a pre-hook in each unit:

- name: cert-manager-issuer
  type: kubernetes
  source: ./cert-manager/
  provider_version: "0.6.0"
  config_path: ../kubeconfig_{{ .name }}
  depends_on: this.cert-manager
  pre_hook:
    command: *getKubeconfig
    on_destroy: true
    on_plan: true