Good point, but in production environments you'd want to mantain the storage on a separate nodes rather than the ones where you have your deployment, because in case of the node failure your storage is affected and not persisted as well.