Storage

Storage in VS-HCI comes in layers, and the one you choose for a VM’s disks decides whether that VM can move between hosts.

The choice that matters

Backing	Reachable from	VM can migrate
A pool on one host	That host only	No
A replicated storage image	Every host	Yes
An iSCSI LUN	Every host with a session	Yes

A VM on local disk is pinned to its server. It cannot live-migrate, and high availability cannot restart it elsewhere, because its disk exists in one place. That is fine for something you can rebuild and wrong for anything you cannot.

See Live migration.

Storage modes

Three ways to provide shared storage, chosen per deployment:

Replicated storage — Ceph

Three or more converged nodes pool their disks into a single cluster. Data is replicated across nodes, every host can reach every image, and losing a node loses no data.

This is the normal choice for VS-HCI, and the one that makes the platform hyperconverged rather than a set of servers with local disks.

See Replicated storage.

A replicated pair

Two storage nodes mirroring each other, with a storage VIP that follows whichever node is active. Failover is managed by a cluster resource manager rather than by the platform.

Suited to a deployment that has exactly two storage nodes and needs them redundant.

Warning — A replicated pair requires working fencing, and the platform verifies it before making any change. No fencing, no cluster. Two nodes that can both believe they are primary will corrupt the data they are protecting.

Standalone

Local pools on each host, no replication. Appropriate for a single-node deployment, for scratch, and for workloads you can rebuild.

Preflight

Before any storage mode is built, a preflight probe examines the hardware and reports whether it is fit. Each mode has its own gates, and an unfit deployment is refused with a message saying what is wrong.

The intent is to fail at onboarding rather than during a failover at three in the morning. Run it from a host’s storage screen.

Pools

A pool is storage on one host that VM disks live in.

Type	What it is
DIR	A directory on the host’s filesystem
LVM	An LVM volume group
ZFS	A ZFS pool
NFS	A mounted NFS export
ISCSI	An attached iSCSI target
CEPH	A replicated storage pool

Pool status is ACTIVE, INACTIVE, BUILDING, DEGRADED or ERROR. See Pools.

Exports

The platform can also serve storage: iSCSI targets exported from storage nodes, consumed by hosts and by machines outside the deployment. One VM disk is one LUN.

See Exports.

The ISO library

One library the whole fleet sees at the same path, so a VM definition naming an ISO stays valid wherever the VM runs. See The ISO library.

Storage on its own network

By default, storage would share the management link — and a management problem would become a storage outage. A storage distributed switch gives every host a second address and moves storage traffic onto it.

Replication is worth separating again: a rebuild running at line rate for hours degrades the storage it is rebuilding. A replication switch on a dedicated uplink puts it on different wires entirely.

See Distributed switches.

Where to look

Question	Page
What pools does this host have?	Pools
How do I add a disk to a VM?	Volumes
How do I build shared storage?	Replicated storage
How do I serve storage out?	Exports
Where do ISOs live?	The ISO library