Your work needs more than a disk.

A server rarely starts as a clean specification. It starts with the disks already on the shelf, an application someone needs this week, and another tool added when the last one reaches its limit. Storage, file sharing, containers and virtual machines end up managed through different interfaces. The owner becomes the integration layer.

Storage makes that fragmentation expensive. A creator's active work and a second box of replaceable media do not have the same requirements. Building both around one storage policy either wastes hardware or hides a compromise. We wanted the interface to explain the choice instead of making it for everyone.

It is for creators who need their work backed up, home server owners growing beyond a collection of containers, and small businesses that need shared storage without handing every user the keys to the whole machine. Different data deserves different decisions: important working files on ZFS, an archive of old media on a pool of mixed drives, and a separate backup destination for recovery.

We brought storage pools, file shares, Docker applications, virtual machines and system containers into one web interface. Nova, the built-in assistant, adds a way to investigate the system in ordinary language. The app catalog extends the machine without making every owner assemble deployment files from scratch.

We also built the parts between a download and a usable NAS: installation, network setup, first-array creation, permissions and updates. The storage wizard belongs in first-run setup because a NAS with blank disks is not ready for anyone's work.

For important data, ZFS provides the integrity and snapshot capabilities that justify planning the drives together. For a media archive that grows one sale-priced disk at a time, MergerFS presents mixed drives as one pool and SnapRAID adds scheduled parity. Each file remains whole on an underlying disk. Files written since the last parity sync are not protected by that sync, so this is a deliberate archive choice.

Backups belong on another destination. Eight.ly Backup Server supplies the deduplicating backup target, while the OS brings the management surfaces together. The recovery workflow has to account for the data, configuration and applications a creator needs to get back to work.

The difficult work was making the interface agree with the machine. ZFS availability now depends on the actual kernel module, not merely the presence of command-line tools. Permission failures remove access rather than defaulting to full control. Self-updates verify the replacement, check its health and have a rollback path.

Failure testing also exposed limits. In a July 2026 three-node power-loss drill, the system detected the missing node but failover did not complete because the saved fencing configuration never reached the engine. We recorded that failure instead of calling it high availability. Infrastructure earns trust by showing what happened, including the part that did not work.