Currently I am running a single FoundationDB Kubernetes cluster on GKE with the FDB operator. Due to performance sensitive workloads I use nodepools with local SSDs. One headache that arises from using local SSDs, is that every time that I need to upgrade the Kubernetes version on the nodepool, all the data stored on the local SSDs is lost. So what would be the most optimal workflow to follow when upgrading the Kubernetes cluster with the least pain endured? Should I just exclude an instance one by one, upgrade it and then insert it back while using a lower replication factor to speed things up? Would it be possible to migrate to a new nodepool with the latest Kubernetes version and transfer the data there?