preStop.lifecycle.sleep, with helm chart scales down deployment to 1 replica

[mclarke47]

Checklist:

• I've searched in the docs and FAQ for my answer: https://bit.ly/argocd-faq.
• I've included steps to reproduce the bug.
• I've pasted the output of argocd version.

Describe the bug

When using:

lifecycle:
  preStop:
    sleep:
      seconds: 120

With ArgoCD and a helm chart as a source, ArgoCD will continually override the deployment forcing it to scale to 1 replica, regardless of HPA configuration.

To Reproduce

1. Create a basic helm chart with helm init and turn autoscaling.enabled: true and autosclaing.minReplicas: 2 apply the application

2. Update the chart to use

lifecycle:
  preStop:
    sleep:
      seconds: 120

wait for the application to sync

Note the deployment has been scaled down to 1 pod (despite the HPA minReplicas), it will repeatedly try to scale up but new pods will be deleted.

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: sleep-test
  namespace: argocd
spec:
  project: default
  destination:
    server: https://kubernetes.default.svc
    namespace: default
  source:
    chart: sleep-test
    helm:
      values: |
        autoscaling:
          enabled: true
          minReplicas: 2
      version: v3
    repoURL: repo-for-the-chart
    targetRevision: 0.1.0
  syncPolicy:
    automated:
      selfHeal: true
    syncOptions:
    - Replace=true
    - Force=true
    - Prune=true

Expected behavior

This lifecycle setting should not interfere with scaling.

Screenshots

Version

argocd-server: v2.14.7+d107d4e
  BuildDate: 2025-03-19T19:56:45Z
  GitCommit: d107d4e41a9b8fa792ed6955beca43d2642bad26
  GitTreeState: clean
  GoVersion: go1.24.1
  Compiler: gc
  Platform: darwin/arm64
  Kustomize Version: v5.6.0 2025-01-14T15:08:34Z
  Helm Version: v3.17.1+g980d8ac
  Kubectl Version: v0.31.0
  Jsonnet Version: v0.20.0