🧩 Problem Overview

• Context: Using Argo CD to manage ingress-nginx Helm chart.

• Issue Introduced: In chart version 4.12.3+, pre-sync and post-sync Jobs have ttlSecondsAfterFinished: 0, meaning the Job itself is deleted immediately after completion.

• Symptom: Argo CD sync remains stuck in Pending or Waiting—Jobs finish, Kubernetes garbage‑collects them immediately, but Argo CD can't track completion and retries indefinitely.

🔍 Root Cause

The chart’s hardcoding leads to:

1. Job runs and completes.

2. ttlSecondsAfterFinished: 0 triggers deletion of the Job object.

3. Argo CD sees the Job’s absence on the cluster → marks it OutOfSync, re-creates it → loop.
   Similar behavior confirmed in community answers: Argo CD re-syncs Jobs removed due to ttlSecondsAfterFinished.

✅ Workarounds

🟢 1. Use Chart Version 4.12.2 + Patch

• Deploy ingress-nginx v4.12.2 to bypass ttl issue.

• Still use the latest container image by patching image tag & digest via values.yaml:

  controller:
    image:
      tag: <latest-tag>
      digest: <latest-digest>
  

• This avoids introducing ttl bug while benefiting from newest controller releases.

🛠️ Step-by-Step Troubleshooting

1. Confirm Chart Version & Behavior

   • Deploy v4.12.3+ with Argo CD → observe Jobs disappear quickly, sync never completes.

   • kubectl get jobs after sync stuck.

2. Validate Argo CD Logs

   • argocd-server logs indicate re-application of the Job resource.

3. Implement Fix

   • Pin chart to 4.12.2 + patch image.

4. Test and Confirm
   

   • Trigger Re-sync on ArgoCD UI. It should complete without stuck Jobs.

   • Validate no lingering Jobs and Application health is green.

5. Monitor Upstream Changes

   • Once ttlSecondsAfterFinished is made configurable and published, re-adopt latest chart version and remove patch.

📌 Sample Patch via Helm values.yaml

controller:
  image:
    tag: "v1.8.1" # or current latest
    digest: "sha256:<digest>"

Ensure chart dependency still references 4.12.2 in Chart.yaml.

🎯 Key Takeaways