OviOS never has to be installed. Boot it from a USB stick, an ISO or the network and it runs entirely from RAM. Your whole configuration lives on your ZFS pools and comes back by itself at every boot. Install it to disk only if you want to.
The operating system is disposable; your configuration is not. OviOS 6.2 keeps the complete node configuration on your storage pools, so any boot medium becomes a working storage controller in minutes, identical to the one before it.
Start from USB, ISO or PXE with ovios.autorestore=1. OviOS loads into RAM. Nothing is
installed and no system disk is needed.
Create pools, LUNs, targets, shares and users as usual. Every change is saved automatically to a hidden dataset on every pool, as one of 10 rolling snapshot generations.
Power-cycle, swap the USB stick or boot a fresh image. Before any service starts, OviOS imports the pools and restores the newest generation, including iSCSI, NFS, SMB and all options.
Every drive bay goes to data, and there is no OS disk to wear out, patch in place or back up. A dead USB stick costs you minutes, not a rebuild.
A copy of the configuration on every pool, verified by checksum, with rollback across generations. As long as one pool imports, the node knows exactly what it was.
Upgrade running nodes one at a time, whether installed or running live from RAM, without taking storage offline. The configuration carries across releases. On a live node you can also simply boot a newer image.
Each node is recognised by its DMI UUID, with a stable ZFS hostid derived from it, so several nodes can share pools without picking up each other's configuration.
With smb.secrets.restore on, SMB passwords and the AD join travel with the configuration.
Leave it off and those secrets never leave the node.
Installing to disk (ext4 or ZFS root, UEFI or BIOS) is still fully supported. Installed systems can be
stateless too: options auto.restore.enable on.
Stateless operation leads the release, but there is a lot more. Here are the highlights; the release notes have everything.
No installation needed. The configuration is saved to every pool and restored at boot, with node identity
from the DMI UUID and the new smb.secrets.restore option.
Incoming, outgoing and mutual CHAP per target from the new target chap menu, with secrets
kept out of command lines and logs.
Space freed by clients (Linux fstrim, Windows ReTrim, VMware VMFS6) goes back to the pool automatically.
One option turns on MinIO-based S3, and a wizard sets up the volume, credentials and certificates.
HTTPS by default and CSRF, XSS and injection fixes. Errors are shown, and the version is displayed everywhere.
Atomic config sync between partners, safer disk handling, and fixes across snapshots, volumes and LUNs.
Everything a storage OS should do, and nothing it shouldn't.
Advanced data integrity, RAID-Z, and snapshots built into the core OS architecture.
Built-in clustering tools for automatic failover and mission-critical uptime.
*Contact for HA support.Stripped of non-essential services to ensure every CPU cycle goes to storage I/O.
A menu-driven CLI environment for managing pools, shares, and services without memorizing commands.
Enterprise-tuned support for iSCSI, NFS, SMB, and options like S3 and NVMe over Fabrics (TCP + RDMA) from one unified stack.
The whole OS runs from memory, leaving all disk bandwidth free for data. No installation needed; the configuration is restored from your pools at every boot.