k8s - allow resource definition using generateName

[abikouo]

SUMMARY

#35

ISSUE TYPE

• Feature Pull Request

COMPONENT NAME

k8s

ADDITIONAL INFORMATION

- name: create pod using generateName
  k8s:
    namespace: test
    generate_name: pod-
    definition:
       kind: Pod
       spec:
          containers:
          - name: py
            image: python:3.7-alpine

- name: create pod using generateName
  k8s:
    namespace: test
    definition:
       kind: Pod
       metadata:
          generateName: pod-
       spec:
          containers:
          - name: py
            image: python:3.7-alpine

[gravesm]

There are a few problems that will need to be sorted out. In general, we'll need to go through very carefully and make sure that in every instance that we are accessing, or trying to access, the name that we are accounting for the changes here. Some examples of things that are not working, all for different reasons:

k8s:
  kind: Namespace
  generate_name: ns-

k8s:
  kind: ConfigMap
  namespace: whatever
  generate_name: cmap-
  append_hash: yes

k8s:
  definition:
    apiVersion: v1
    kind: Pod
    metadata:
      namespace: whatever
      name: pod1
    spec:
      containers:
        - image: adslfkjadslfkjadslkfjsadf
          name: non-existent-container

k8s:
  definition:
    apiVersion: v1
    kind: Pod
    metadata:
      namespace: whatever
    spec:
      containers:
        - image: busybox
          name: c1
  generate_name: pod-
  wait: yes

In the latter case, because there is no name, we end up waiting on every pod in that namespace. This wait will fail because the first pod has a non-existent image name.

[abikouo]

There are a few problems that will need to be sorted out. In general, we'll need to go through very carefully and make sure that in every instance that we are accessing, or trying to access, the name that we are accounting for the changes here. Some examples of things that are not working, all for different reasons:

k8s:
  kind: Namespace
  generate_name: ns-

k8s:
  kind: ConfigMap
  namespace: whatever
  generate_name: cmap-
  append_hash: yes

k8s:
  definition:
    apiVersion: v1
    kind: Pod
    metadata:
      namespace: whatever
      name: pod1
    spec:
      containers:
        - image: adslfkjadslfkjadslkfjsadf
          name: non-existent-container

k8s:
  definition:
    apiVersion: v1
    kind: Pod
    metadata:
      namespace: whatever
    spec:
      containers:
        - image: busybox
          name: c1
  generate_name: pod-
  wait: yes

In the latter case, because there is no name, we end up waiting on every pod in that namespace. This wait will fail because the first pod has a non-existent image name.

ok for the latter case, I see the issue with the wait mechanisms, I need to check for other scenarios
Thanks @gravesm

[abikouo]

@gravesm regarding the latter case you just mentioned in your comment above, it is not failing the resource name is updated before calling the wait method,

kubernetes.core/plugins/module_utils/common.py

Lines 824 to 825 in 2ffe813

	definition['metadata'].update({'name': k8s_obj['metadata']['name']})
	success, result['result'], result['duration'] = self.wait(resource, definition, wait_sleep, wait_timeout, condition=wait_condition)

I have added test case to validate that

kubernetes.core/molecule/default/tasks/generate_name.yml

Lines 149 to 180 in 2ffe813

	- name: Create Pod with failing container
	kubernetes.core.k8s:
	namespace: '{{ namespace }}'
	definition:
	apiVersion: v1
	kind: Pod
	metadata:
	name: pod1
	spec:
	containers:
	- image: adslfkjadslfkjadslkfjsadf
	name: non-existent-container-image
	- name: Create second Pod using wait (it should not wait for the first container)
	kubernetes.core.k8s:
	namespace: '{{ namespace }}'
	definition:
	apiVersion: v1
	kind: Pod
	metadata:
	name: pod2
	spec:
	containers:
	- args:
	- /bin/sh
	- -c
	- while true; do echo $(date); sleep 10; done
	image: python:3.7-alpine
	imagePullPolicy: Always
	name: c0
	wait: yes
	wait_timeout: 10

For others scenarios, code has been fixed and test updated accordingly

[abikouo]

@Akasurde could you please take a look at this one?
Thanks

[ansible-zuul]

LGTM!