ArgoCD detects constant drift on StatefulSet volumeClaimTemplate

[gianvetter]

Bug Description

ArgoCD continuously detects drift and attempts to sync StatefulSet resources (alertmanager, compactor, ingester, store-gateway) due to missing apiVersion and kind fields in volumeClaimTemplates.

Problem

When deploying the Cortex Helm chart, Kubernetes automatically populates the apiVersion: v1 and kind: PersistentVolumeClaim fields in the StatefulSet's volumeClaimTemplates during resource creation. However, the Helm templates don't include these fields, causing ArgoCD to detect a constant difference between:

• Desired state (from Helm): volumeClaimTemplates without apiVersion/kind
• Actual state (in cluster): volumeClaimTemplates with apiVersion/kind

This results in ArgoCD sync flapping, constantly trying to "fix" a difference that reappears immediately.

Expected Behavior

StatefulSet templates should explicitly include apiVersion and kind fields in volumeClaimTemplates to match the actual cluster state and prevent ArgoCD drift detection.

Affected Components

• templates/alertmanager/alertmanager-statefulset.yaml
• templates/compactor/compactor-statefulset.yaml
• templates/ingester/ingester-statefulset.yaml
• templates/store-gateway/store-gateway-statefulset.yaml

Reference

The Grafana Loki Helm chart uses this pattern successfully: https://github.com/grafana/loki/blob/main/production/helm/loki/templates/ingester/statefulset-ingester.yaml#L83-L85