This is where complexity comes in. Let Kubernetes abstract the resources for your workloads. For example, use a storage service (like Rook/openEBS/portworx/etc

This is where complexity comes in. Let Kubernetes abstract the resources for your workloads. For example, use a storage service (like Rook/openEBS/portworx/etc) that uses node-attached disks to create a data layer. Then you can assign volumes to your individual workloads created and managed by the storage service. This provides reliable and portable storage for your other services without having to tying them to lower-level details.