Add a replicas configuration option for kustomize

[toyamagu-2021]

Summary

Currently, kustomize configuration options are available only for namePrefix, nameSuffix, images, commonLabels, commonAnnotations according to argocd-doc.

I would like to see a replicas transformer to change a number of replicas kustomize-doc.

Motivation

I would like to change the number of replicas in application.yaml. Like,

apiVersion: argoproj.io/v1alpha1
kind: Application
spec:
    # kustomize specific config
    kustomize:
      replicas:
        - name: deployment-name
          count: 5

Proposal

I think it is straightforward.

• Kustomize already implements kustomize edit set replicas deployment-name=5.
• And our code is ready to add the above command in

  argo-cd/util/kustomize/kustomize.go

  Line 87 in b4c1281

  func (k *kustomize) Build(opts *v1alpha1.ApplicationSourceKustomize, kustomizeOptions *v1alpha1.KustomizeOptions, envVars *v1alpha1.Env) ([]*unstructured.Unstructured, []Image, error) {

• So, we can implement this feature by following the same way as fix: adding commonAnnotations #4613

[crenshaw-dev]

Sounds great to me. Want to open a PR?

[toyamagu-2021]

@crenshaw-dev
I'll gladly do!

[maltewhiite]

I would love to do this with ApplicationSet as well @toyamagu-2021

ApplicationSet Example Yaml

apiVersion: argoproj.io/v1alpha1
kind: ApplicationSet
metadata:
  name: guestbook
  namespace: argocd
spec:
  generators:
  - list:
      elements:
      - namespace: guestbook-qa
        url: https://kubernetes.default.svc
        values:
          frontend:
            replicas: 1
      - namespace: guestbook-e2e
        url: https://kubernetes.default.svc
        values:
          frontend:
            replicas: 2
      - namespace: guestbook-stage
        url: https://kubernetes.default.svc
        values:
          frontend:
            replicas: 3
      - namespace: guestbook-prod
        url: https://kubernetes.default.svc
        values:
          frontend:
            replicas: 5
  template:
    metadata:
      name: '{{namespace}}'
    spec:
      project: gitops
      source:
        repoURL: https://gitlab.com/kubernetes-practice/my-k8s/kia.git
        targetRevision: HEAD
        path: gitops-and-kubernetes/guestbook
        kustomize:
          replicas:
            - name: frontend
              count: '{{values.frontend.replicas}}'
      destination:
        server: '{{url}}'
        namespace: '{{namespace}}'
      syncPolicy:
        syncOptions:
        - CreateNamespace=true
        automated:
          prune: true
          selfHeal: true

[toyamagu-2021]

@maltewhiite
Thanks for the detailed suggestions.
Yes, we can do that after #11665 is merged. I used the following manifest for the test:

applicationset.yaml

apiVersion: argoproj.io/v1alpha1
kind: ApplicationSet
metadata:
  name: guestbook
  namespace: argocd
spec:
  generators:
  - list:
      elements:
      - namespace: guestbook-qa
        url: https://kubernetes.default.svc
        values:
          replicas: "1"
      - namespace: guestbook-e2e
        url: https://kubernetes.default.svc
        values:
          replicas: "2"
      - namespace: guestbook-stage
        url: https://kubernetes.default.svc
        values:
          replicas: "3"
      - namespace: guestbook-prod
        url: https://kubernetes.default.svc
        values:
          replicas: "5"
  template:
    metadata:
      name: '{{namespace}}'
    spec:
      project: default
      source:
        repoURL: https://github.com/argoproj/argocd-example-apps.git
        targetRevision: HEAD
        path: kustomize-guestbook
        kustomize:
          replicas:
            - name: kustomize-guestbook-ui
              count: '{{values.replicas}}'
      destination:
        server: '{{url}}'
        namespace: '{{namespace}}'
      syncPolicy:
        syncOptions:
        - CreateNamespace=true
        automated:
          prune: true
          selfHeal: true

Please let me know if there is any misunderstanding.

[maltewhiite]

@toyamagu-2021 Looks great! Thank you very much. Ideally my team would use auto-scalers, but this will help for sure.