Storage From Business
Distributed storage deployment and management from the same panel that runs your VMs, office PCs and websites: build the cluster, add disks and hosts, give each owner a quota, take snapshots and roll back — and see which workload is using which storage.
On this page: Cluster · Data services · Who uses it · Capacity · Snapshots · Limits
Container-based orchestrated deployment; new clusters start on version 20.2.4.
Add hosts and disks; group disks by label so pools land on the disks you intend. Storage pool expansion means adding disks or hosts.
Orchestrated version upgrades (19.2.6 → 20.2.4 tested).
Stretch mode with two rooms plus an arbiter has been tested.
| Service | What it is used for in Nodekeeper |
|---|---|
| Block devices | VM disks, office PC system disks, members' block storage — with quotas and permissions |
| Distributed file storage | Website storage, members' “My storage” subvolumes, the CDN shared cache |
| S3 object storage | S3-compatible buckets on your own cluster |
| Multiple active metadata servers | File storage metadata scale-out |
| NVMe-oF/TCP gateway | Block storage for clients that speak NVMe over TCP |
| Replicas / erasure coding | Replicas or erasure coding 2+1, 2+2, 4+2 and 8+3 |
Cloud PC system disks are block devices on distributed storage; distributed file storage holds website storage and members' 'My storage'. Each PC has its own writable disk cloned from a shared template version. Office PCs
KVM and Proxmox VE disks on distributed block storage, which is what lets a VM restart on another host. High availability
Site files on distributed file storage and per-member storage, under the same isolation rules as local sites. Web hosting
Shared base image + per-PC differential: a template version is a protected snapshot, and each PC is a copy-on-write clone of it, so space is used by what each PC actually writes, not by the full size of every disk. Capacity on demand uses thin-provisioned disks and a quota per owner.
When an owner reaches the quota: warning, then a 24-hour grace period, then writes stop. Plan four numbers separately — logical size handed out, space actually used, replicas or erasure-coding overhead, and the reserve you keep free.
Snapshots are fast to create, crash-consistent (no in-guest agent), and live inside the same cluster: they are recovery points, not backups. Keep independent copies with a backup job.
Distributed storage is part of Business. The trial covers up to 3 nodes; talk to us before you size a larger cluster.