Do not set .spec.replicas for store-gateway if autoscaling is enabled

[alex-berger]

If a HorizontalPodAutoscaler (HPA) is in place, one must not explicitly set .spec.replicas on the corresponding workload (StatefulSet or Deployment). Failing to adhere to that rule can cause severe service outage as the workload is scaled down unexpectedly and then might crash as it immediately gets overloaded (RAM or CPU).

In https://github.com/cortexproject/cortex-helm-chart/blob/master/templates/store-gateway/store-gateway-statefulset.yaml#L13

  replicas: {{ .Values.store_gateway.replicas }}

should actually be:

  {{- if not .Values.store_gateway.autoscaling.enabled }}
  replicas: {{ .Values.store_gateway.replicas }}
  {{- end }}

[nschad]

Thanks.

Do you want to give it a go?

[husnialhamdani]

Hi @nschad,
I Interested to work on it, can you please assign to me?