Nearly Full CSV on S2D Cluster, but Guest VHDX Shows 50% Used

Question

Environment:

• Windows Server 2019 Datacenter, 4‑node HCI/S2D cluster
• The volume disk that has the symptoms is 12 TB (“ClusterDisk01”)
• Hyper‑V VMs with fixed VHDX (maximum size set to a maximum of 12 TB)

Symptoms:

• Failover Cluster Manager reports CSV “ClusterDisk01” almost 95% utilized.
• Inside the running VM on that CSV, C: drive (VHDX) shows less than that almost 6 TB.
• No large local VM snapshots

Questions:

• Are these CSV cache or logs? and what best practices exist to manage their growth?
• Can I safely reduce the CSV cache reservation to reclaim space without impacting performance?
• What is the recommended way to shrink fixed VHDX files in a production cluster?
• Are there known issues with ReFS metadata overhead on large S2D CSVs?

Any guidance or documentation pointers would be greatly appreciated—thanks in advance!