Let's imagine you have an EKS cluster with a Grafana deployment. You cluster has as a default storage EBS. Unfortunately EBS doesn't allow you to create more replicas of your deployment because the EBS is already attached to the instance. To workaround this we can use EFS.
Before you start changing your workloads you install Amazon EFS CSI driver as an addon.
Then create an EFS disk and write down the identifier. It's better to have an EFS per deployment (better maintenance for example).
Now we have a possibility to create a StorageClass in our Helm chart:
kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
name: efs-storageclass
provisioner: efs.csi.aws.com
parameters:
provisioningMode: efs-ap
fileSystemId: {{ .Values.grafana_custom.StorageClass.efs_id | quote }}
directoryPerms: "755"
basePath: "/grafana"
When a PVC is created with a StorageClass that has a basePath specified, the dynamically provisioned PV will be created with a subdirectory path under the specified basePath. This can be useful for organizing and managing the storage resources in a cluster.
It is better to first deploy the StorageClass before you change the Grafana storage type.
Change the StorageClass name in the Helm values: https://github.com/grafana/helm-charts/blob/main/charts/grafana/values.yaml#L327
Change the access mode to "ReadWriteMany": https://github.com/grafana/helm-charts/blob/main/charts/grafana/values.yaml#L328
Enter the needed disk size: https://github.com/grafana/helm-charts/blob/main/charts/grafana/values.yaml#L329
