`config.logs.enabled=true` when `config.jobs.deploy-k8s-false` is broken

[viccuad]

Using the following config file:

(note logs.enabled=true and jobs.deploy-k8s=false)

workertags:
  caasp4: suse-internal

jobs:
  deploy-k8s: false
  deploy-kubecf: true
  smoke-tests: true
  cf-acceptance-tests-brain: false
  sync-integration-tests: false
  minibroker-integration-tests: false
  cf-acceptance-tests: false
  upgrade-kubecf: false
  deploy-stratos: false
  destroy-kubecf: true
jobs_ordered:
- deploy-k8s
- deploy-kubecf
- smoke-tests
- cf-acceptance-tests-brain
- sync-integration-tests
- minibroker-integration-tests
- cf-acceptance-tests
- upgrade-kubecf
- deploy-stratos
- destroy-kubecf
backends:
  caasp4: true
  aks: false
  gke: false
  eks: false
availabilities:
  sa: true
  ha: false
  all: false
schedulers:
  diego: true
  eirini: false
catapult:
  uri: https://github.com/SUSE/catapult
  branch: master
brain:
  verbose: false
  inorder: false
  include: ""
  exclude: ""
cats:
  nodes: 3
  flake-attempts: 5
  timeout-scale: "3.0"
s3minibroker:
  bucket: minibroker-helm-charts
  region: us-east-1
  regexp: minibroker-charts/minibroker-(.*).tgz
s3:
  bucket: kubecf
  region: us-west-2
  regexp: kubecf-bundle-v(.*).tgz
# Scheduling is required for nightly builds, enabling it removes regular trigger on kubecf bundle.
schedule:
  enabled: false
  start: 12:00 AM
  stop: 12:10 AM
  location: America/Vancouver
logs:
  enabled: true

Gives me:

apply configuration? [yN]: y
error: invalid pipeline config:
invalid resources:
        resource 's3.klog-destination' is not used

[viccuad]

This is happening because the klog script is only called inside the destroy-k8s task:

cap/cap-ci/common/destroy_k8s.tmpl

Lines 40 to 52 in 515c930

	{{- if eq .logs.enabled true }}
	{{ tmpl.Exec "scripts_klog" | indent 14 | trimSpace }}
	{{ end }}
	{{ tmpl.Exec "scripts_destroy_k8s" | indent 14 | trimSpace }}
	- put: {{ .backend }}-pool.kube-hosts
	params:
	remove: {{ .backend }}-pool.kube-hosts
	depth: 1
	{{- if eq .logs.enabled true }}
	- put: s3.klog-destination
	params:
	file: catapult/klog-*.tar.gz
	{{ end }}

Yet (rightly) the definition of the resource has nothing to do with destroy_k8s:

cap/cap-ci/pipeline.yaml.tmpl

Lines 88 to 98 in 515c930

	{{- if $config.logs.enabled }}
	- name: s3.klog-destination
	type: s3
	source:
	bucket: kubecf-klog
	access_key_id: ((aws-access-key))
	secret_access_key: ((aws-secret-key))
	regexp: klog-(.*)\.tar\.gz$
	region_name: us-west-2
	endpoint: null
	{{ end }}

Possible solutions:

• IMHO, the correct solution would be to create another task klog analogous to destroy_k8s, that is run on_failure if enabled. This means calling a common_klog just before all common_destroy_k8s, if klog is enabled. Gave a quick try in https://github.com/SUSE/cap/compare/viccuad/fix-klogs-destroy, but fails so far. See commit messages.

• A cheap, spaghetti solution would be: (pseudocode) {{- if ($config.logs.enabled or $jobs.deploy-k8s == false) }} in https://github.com/suse/cap/blob/master/cap-ci/pipeline.yaml.tmpl#L88

[prabalsharma]

There is a fundamental issue of design here. klog should not be part of destroy k8s job (aka destroy-kubecf flag) https://jira.suse.com/browse/CAP-1564