ArgoCD-related issues and solutions
A quick reference for ArgoCD problems I have hit and the fix that worked.
| Symptom | Cause | Fix |
|---|---|---|
kubectl port-forward drops or breaks after a few minutes | Idle connection timeout | Keep the tunnel alive with a looped nc probe |
failed to replace object: Service "argocd-server" | Helm chart is forcing an immutable service update | Remove force_update = true from the Helm release |
server.secretkey is missing | ArgoCD server secret was reset or not initialized | kubectl rollout restart deploy/argocd-server -n argocd |
error getting cached app resource tree: cache: key is missing | Application controller cache is stale | kubectl rollout restart statefulset -n argocd argocd-application-controller |
Keeping a port-forward alive
kubectl port-forward can drop when the connection goes idle. Open a second terminal and keep probing the local port:
while true; do nc -vz 127.0.0.1 8080; sleep 10; done
Replace 8080 with the port you forwarded.
Completion criterion
The ArgoCD issue is resolved when:
- The UI or CLI is reachable and stays reachable.
- The application resource tree loads without cache errors.
- Any service-related error is gone after removing
force_update = true.