WeRecoverData

Nutanix Data Recovery

Nutanix clusters run their storage in software: every node's local disks are pooled by a Controller VM into one distributed storage fabric shared by all nodes. When that fabric is damaged — nodes removed too quickly, metadata inconsistency, a cluster destroyed during maintenance — the virtual machines become unreadable even though the physical disks are intact.

Platform Lineage and Naming

Nutanix shipped its first hyperconverged appliances in 2011. The storage layer was originally called the Nutanix Distributed File System (NDFS), then the Distributed Storage Fabric (DSF), and is now AOS Distributed Storage — the same fabric under successive names.

Nutanix Unified Storage (NUS) is a separate product grouping layered on top: Nutanix Files (SMB/NFS file services), Nutanix Objects (S3 object storage) and Nutanix Volumes (iSCSI block services). NUS services consume AOS Distributed Storage underneath rather than replacing it — so a NUS case is normally an AOS recovery plus the service's own structures.

Generations and Models We Evaluate

Architecture and Data Layout

Each node runs a Controller VM that owns the node's local disks and joins them into a cluster-wide storage pool. Guest virtual disks (vDisks) are broken into extents, grouped into extent groups, and replicated across nodes according to the redundancy factor — the physical placement of a VM's data is spread across the cluster, not held on its host.

A distributed metadata service (built on a Cassandra-derived ring) maps vDisks to extent groups and their replicas. Metadata quorum loss is a defining Nutanix failure mode: the extent data can be intact on disk while nothing can be resolved back into a usable virtual disk.

Nutanix Unified Storage services sit above this fabric. Nutanix Files runs file server VMs whose shares live in AOS-backed volume groups; Nutanix Objects stores object data and its own metadata in AOS; Nutanix Volumes presents iSCSI volume groups directly. Recovering a NUS share or bucket therefore means reconstructing the AOS layer first, then the service's own file system or object metadata.

Reconstruction is performed from images of the drives in every contributing node — rebuilding extent groups, resolving metadata, reassembling vDisks, and then recovering the guest file systems, shares or objects inside them.

Logical Failures

Hardware Failures

Encryption and Keys

Nutanix supports software data-at-rest encryption with a local key manager or an external KMIP key server, and self-encrypting drives on some node types. Key material and cluster key state must be preserved with the drives; encrypted extent groups cannot be interpreted without them.

Frequently Asked Questions

Is Nutanix Unified Storage the same thing as AOS storage?

No. AOS Distributed Storage is the underlying fabric that pools node disks. Nutanix Unified Storage — Files, Objects and Volumes — is a set of services that run on top of that fabric. Recovering a NUS share or bucket normally requires reconstructing the AOS layer first.

Our cluster was destroyed in Prism. Can VMs be recovered?

Sometimes. Cluster destroy operations remove configuration and metadata, but extent data frequently remains on the drives. Stop all further cluster creation or re-imaging immediately, since those steps overwrite what a recovery would use.

Do you need every node?

Drives from every node that held replicas of the affected data are normally needed. With RF2 or RF3 a subset can sometimes be sufficient, which the evaluation determines.

Does it matter which hypervisor was used?

The AOS-layer reconstruction is the same for AHV, ESXi and Hyper-V. The hypervisor determines the guest disk format handled at the final extraction stage.

Related