VMware 3V0-23.25 - Advanced VMware Cloud Foundation 9.0 Storage

vSAN Stretched Clusters meet the availability-zone requirement by extending a single logical vSAN cluster across two availability zones and using a witness for quorum. This provides rapid recovery from a site or availability-zone failure because workloads remain protected by the stretched topology and vSAN site-failure tolerance. vSAN Storage Cluster addresses the need for scalable, high-performance storage that can grow independently as application capacity requirements increase. In VCF 9.0, a vSAN Storage Cluster is based on vSAN ESA and provides disaggregated storage to vSAN compute or HCI clusters. This design supports storage-dense NVMe-based nodes, high IOPS, low latency, independent capacity scaling, and centralized storage services. Encryption at rest and in transit are native vSAN capabilities that can be enabled on the vSAN-backed storage architecture. vSAN File Service is for file shares, and vSAN iSCSI Target Service is for presenting block LUNs to external initiators; neither is required by the stated application-storage design. vSAN Data Protection is valuable for snapshots and recovery points, but the core architecture requirements are best satisfied by stretched site resilience plus scalable vSAN storage clustering. Reference topics: vSAN Storage Cluster, vSAN Stretched Cluster, vSAN ESA, vSAN Encryption.

The required action is to deploy the VMware Live Recovery appliance. In VMware Cloud Foundation 9.0, vSAN Data Protection is powered by VMware Live Recovery and provides local virtual machine protection and remote virtual machine replication for supported vSAN ESA environments. The vSAN Services configuration workflow explicitly states that before vSAN Data Protection can be used, the VMware Live Recovery appliance must be deployed. This appliance provides the data protection control plane required for protection groups, native snapshots, recovery workflows, and replication-based protection. Enabling vSAN advanced options is not sufficient because those settings control cluster behaviors such as repair timers, read locality, thin swap, unmap, and rebalance. Data Services Manager is used for database/data-service lifecycle use cases, not vSAN VM data protection. vSphere Replication can be part of broader site-recovery scenarios, but it is not the required action to enable vSAN Data Protection from the vSAN services workflow. Reference topics: vSAN Data Protection, VMware Live Recovery Appliance, vSAN Services Configuration, Local VM Protection and Replication.

The direct action is to convert the latency-sensitive OLTP workload policy from RAID-5/6 erasure coding to RAID-1 mirroring while keeping FTT=1. RAID-5/6 provides space efficiency, but it introduces parity operations and additional backend I/O, which can increase write latency during peak transactional activity and resynchronization. VMware states that RAID-1 mirroring requires fewer I/O operations to the storage devices and can provide better performance; cluster resynchronization also completes faster with RAID-1. This directly addresses the observed symptom: resync I/O and backend congestion increasing latency despite adequate network bandwidth. Increasing Object Space Reservation to 100% changes capacity preallocation but does not remove parity overhead or reduce write amplification from erasure coding. Increasing FTT to 2 with RAID-5/6 would increase protection overhead and backend work, making latency worse. Disabling checksums is not valid for ESA because object checksum is always enabled and cannot be deactivated. For ESA, compression can be controlled by policy, but the most direct protection-compliant performance action is RAID-1 FTT=1 for OLTP objects. Reference topics: vSAN Storage Policies, RAID-1 Mirroring, RAID-5/6 Erasure Coding, vSAN ESA Checksum, Resync and Congestion Metrics.

Four hosts are required to meet the 12 TB usable-capacity requirement after applying the protection policy, data reduction, reserves, and overhead. Each host contributes 8 TB raw capacity from two 4 TB NVMe devices. Three hosts provide 24 TB raw capacity, but once host rebuild reservation, 10% operational reserve, RAID-5 protection overhead, and 25% filesystem/object overhead are considered, the remaining effective usable capacity falls below the required 12 TB target. Four hosts provide 32 TB raw capacity. With vSAN ESA and FTT=1/RAID-5, vSAN uses erasure coding to improve capacity efficiency compared with mirroring. The vSAN sizing guidance states that RAID-5 with FTT=1 provides approximately 75% usable capacity before other reserves and overhead. Host Rebuild Reserve must also be planned so the cluster has enough free capacity to rebuild components after a host failure. After applying the expected 1.5 data reduction ratio and the required reserves, four hosts provide enough usable capacity while three hosts do not. Reference topics: vSAN ESA Capacity Sizing, RAID-5 Erasure Coding, Host Rebuild Reserve, Operations Reserve, vSAN Object Overhead.

vSAN ESA Storage Clusters are the optimal choice because the workload profile is capacity-heavy and benefits from disaggregated storage. The requirement for 2.5 PB of application data, archival workloads, gateways, and development/test applications indicates a need to scale storage capacity independently from compute resources. VMware Cloud Foundation describes a vSAN storage cluster as a disaggregated storage solution based on vSAN ESA. It provides storage resources but not compute resources, and its datastore can be mounted by vSAN compute clusters or vSAN HCI clusters. This model is designed for storage-dense environments where capacity and performance can be expanded without adding unnecessary compute. A standard vSAN ESA HCI cluster is excellent for high-performance workloads, but it scales compute and storage together. vSAN OSA all-flash and hybrid use the older disk-group architecture and are less suitable for this large-capacity, storage-centric design. Therefore, the principal storage recommendation should be vSAN ESA Storage Clusters to meet scalability, performance, and capacity growth requirements. Reference topics: vSAN Storage Cluster, Disaggregated Storage, vSAN ESA, Storage Models.

Full Clone from Volume Snapshot

4kn Support

Online Volume Expansion

Offline Volume Expansion

Static Persistent Volume

VMware Kubernetes Service persistent volumes use vSphere storage integrations such as CNS and pvCSI to provide Kubernetes storage services backed by vSphere datastores and storage policies. The supported persistent volume capabilities include Static Persistent Volume, which allows pre-existing storage to be consumed by Kubernetes workloads; Offline Volume Expansion and Online Volume Expansion, which allow persistent volumes to be resized depending on workload and filesystem state; Full Clone from Volume Snapshot, which enables a new volume to be created from an existing snapshot; and 4kn Support, which supports modern 4K native storage devices where the underlying stack supports them. Storage vMotion with attached Persistent Volumes is not supported because moving an attached Kubernetes persistent volume while it is in active use can break attachment and consistency assumptions. Storage DRS is also not a supported persistent volume capability in this VKS context. Flash Read Cache is not part of the supported VKS persistent-volume capability set. Reference topics: VMware Kubernetes Service Storage, CNS, pvCSI, Persistent Volume Capabilities, Volume Expansion, Volume Snapshot Clone.

Run the Python reboot helper script only on one ESXi host to recover the cluster.

Enable cluster member updates from vCenter Server on all ESXi hosts.

Exit all hosts from maintenance mode.

During a manual vSAN cluster restart after a planned outage, the hosts must first be brought back and exited from maintenance mode so the cluster can begin recovery. The vSAN reboot helper recovery script must then be run on only one ESXi host. Running the script on multiple hosts can create race conditions and unexpected cluster state. After the recovery script completes, cluster member updates from vCenter Server must be enabled on all ESXi hosts so that normal membership updates resume across the cluster. Enabling updates on only one host is incomplete because all hosts must return to normal cluster-member update behavior. Running the Python reboot helper script on all ESXi hosts is explicitly incorrect because the recovery operation must be initiated from a single host only. Disabling cluster member updates is a shutdown-side activity, not the required restart-side recovery action in this question. Reference topics: Manual vSAN Cluster Restart, Reboot Helper Script, Maintenance Mode Exit, Cluster Member Updates.

vSAN object availability is based on two conditions: an intact replica must be available, and more than 50% of the object’s votes must be accessible. In this scenario, vSAN can locate a full replica and 55% of the votes for the noncompliant objects. Because 55% is greater than the quorum threshold, the objects remain available to the vSAN datastore. A noncompliant status means the VM objects do not currently satisfy the assigned storage policy, but it does not automatically mean the VM is inaccessible. The documentation states that if the operational state remains healthy although the object is noncompliant, the virtual machine can continue using the vSAN datastore. If the cluster contains enough resources and the failure is permanent, vSAN automatically recovers failed components and restores compliance. The VM is marked inaccessible only when a full replica cannot be located or when 50% or fewer votes are available. It becomes orphaned only if the VM namespace, such as the .vmx file, cannot be accessed. Reference topics: vSAN Object Health, Quorum Votes, Noncompliant Objects, Inaccessible and Orphaned VM States.

VMware Live Site Recovery uses VM readiness signals to control recovery sequencing between priority groups. Recovery Priority is used to start and validate higher-priority workloads before lower-priority workloads continue, which is important for application dependency chains such as identity services, databases, middleware, and application servers. A VM being registered or powered on does not prove that the guest operating system is usable. Similarly, final replication completion only confirms storage synchronization, not guest readiness. VMware Tools heartbeat provides a guest-level indication that the operating system has started and is responsive. This makes it the correct mechanism for determining that a recovery priority step has completed and that the recovery plan can continue to the next priority group. The “successfully recovered” state is the result of the recovery workflow, not the signal used to progress between priority groups. Reference topics: VMware Live Site Recovery, Recovery Plans, Recovery Priority, VMware Tools Heartbeat, Application Recovery Sequencing.